Patent Application
20190171871
This disclosure generally relates to computer vision.
Machine learning may be used to enable machines to automatically detect and process objects appearing in images. In general, machine learning typically involves processing a training data set in accordance with a machine-learning model and updating the model based on a training algorithm so that it progressively “learns” the features in the data set that are predictive of the desired outputs. One example of a machine-learning model is a neural network, which is a network of interconnected nodes. Groups of nodes may be arranged in layers. The first layer of the network that takes in input data may be referred to as the input layer, and the last layer that outputs data from the network may be referred to as the output layer. There may be any number of internal hidden layers that map the nodes in the input layer to the nodes in the output layer. In a feed-forward neural network, the outputs of the nodes in each layer—with the exception of the output layer—are configured to feed forward into the nodes in the subsequent layer.
Machine-learning models may be trained to recognize object features that have been captured in images. Such models, however, are typically large and require many operations. While large and complex models may perform adequately on high-end computers with fast processors (e.g., multiple central processing units (“CPUs”) and/or graphics processing units (“GPUs”)) and large memories (e.g., random access memory (“RAM”) and/or cache), such models may not be operable on computing devices that have much less capable hardware resources. The problem is exacerbated further by applications that require near real-time results from the model (e.g., 10, 20, or 30 frames per second), such as augmented reality applications that dynamically adjust computer-generated components based on features detected in live video.
Embodiments described herein relate to machine-learning models and various optimization techniques that enable computing devices with limited system resources (e.g., mobile devices such as smartphones, tablets, and laptops) to recognize objects and features of objects captured in images or videos. To enable computing devices with limited hardware resources (e.g., in terms of processing power and memory size) to perform such tasks and to do so within acceptable time constraints, embodiments described herein provide a compact machine-learning model with an architecture that is optimized for efficiency performing various image-feature recognition tasks. For example, particular embodiments are directed to real-time or near real-time detection, segmentation, and structure mapping of people captured in images or videos (e.g., satisfying a video's frame rate requirements). These real-time computer vision technologies may be used to enable a variety of mobile applications, such as dynamically replacing a video capture of a person with an avatar, detecting gestures, and performing other dynamic image processing related to particular objects (e.g., persons) appearing in the scene.
In addition, particular embodiments described herein provide an efficient method for up-sampling keypoint detections. For example, the output of a keypoint detection process for a particular joint may be in the form of a probability model (e.g., heat map) with regions (e.g., 56×56 grids resolution) that corresponding to regions in the input image. Each region may have a probability value that represents the likelihood of the joint of interest being within that region. To minimize computation time and resources, it may be desirable to limit the resolution of the probability model (e.g., limiting the number of grids). Having a low-resolution model, however, means that the size of each grid may be large, which in turn means that the estimated location of the joint of interest is less precise. To improve precision, particular embodiments first identify n number (e.g., 2 or 3) of the most likely candidate regions where the joint could be located and apply a refinement algorithm (e.g., quadrilateral interpolation) to the patch surrounding each grid (e.g., 3×3, 5×5, or 7×7 patch). The interpolation result of each patch may yield a maximum probability location that is different from any of the probability values associated with the regions in the patch. The maximum probabilities of each patch may then be compared to find the location that has the largest interpolated probability value. The coordinates of the location corresponding to the maximum interpolated probability value could be on a continuous spectrum (i.e., not discrete) and therefore could represent a much more precise location of the joint of interest.
Further embodiments described herein provide an efficient method for improving the accuracy of keypoint detections. As described above, the output of a keypoint detection process for a particular joint may be in the form of a probability model (e.g., heat map) with probability values that each represents the probability of an associated region in in the image containing the joint of interest. A probability model may identify several regions with sufficiently high probability values, making it difficult to discern which of those regions is where the joint of interest is depicted. To improve the accuracy of keypoint detection for joints, particular embodiments may generate and use probability models for bones. For example, a joint probability model for a person's left knee may include two candidate locations (e.g., locations with similarly high probability values), and a joint probability model for the person's left hip may also have two candidate locations. To choose between the candidate locations, the pose-estimation system may use a bone probability model generated for the person's left femur. In particular, for each possible pairing between the left-knee candidate locations and left-hip candidate locations, the system may use the bone probability model to compute the probability of the left femur being located between that pair of candidate locations. The resulting probability scores of the different pairs of candidate locations may be compared, and the pair with the highest score may be used as the final estimated locations of the person's left knee and left hip. In a similar manner, an overall optimization problem may be used to select the entire set of joints (e.g., all 19 joints or a subset thereof) so that the probability computations from the corresponding bone probability models are maximized.
The embodiments disclosed herein are only examples, and the scope of this disclosure is not limited to them. Particular embodiments may include all, some, or none of the components, elements, features, functions, operations, or steps of the embodiments disclosed above. Embodiments according to the invention are in particular disclosed in the attached claims directed to a method, a storage medium, a system and a computer program product, wherein any feature mentioned in one claim category, e.g. method, can be claimed in another claim category, e.g. system, as well. The dependencies or references back in the attached claims are chosen for formal reasons only. However, any subject matter resulting from a deliberate reference back to any previous claims (in particular multiple dependencies) can be claimed as well, so that any combination of claims and the features thereof are disclosed and can be claimed regardless of the dependencies chosen in the attached claims. The subject-matter which can be claimed comprises not only the combinations of features as set out in the attached claims but also any other combination of features in the claims, wherein each feature mentioned in the claims can be combined with any other feature or combination of other features in the claims. Furthermore, any of the embodiments and features described or depicted herein can be claimed in a separate claim and/or in any combination with any embodiment or feature described or depicted herein or with any of the features of the attached claims.
The patent or application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.
Embodiments described herein relate to machine-learning models and various optimization techniques that enable computing devices with limited system resources (e.g., mobile devices such as smartphones, tablets, and laptops) to recognize objects and features of objects captured in images or videos. To enable computing devices with limited hardware resources (e.g., in terms of processing power and memory size) to perform such tasks and to do so within acceptable time constraints, embodiments described herein provide a compact machine-learning model with an architecture that is optimized for performing various image-process tasks efficiently. For example, particular embodiments are directed to real-time detection (including classification), segmentation, and structure (e.g., pose) mapping of people captured in images or videos.
In particular embodiments, the machine-learning model 200 includes several high-level components, including a backbone neural network, also referred to as a trunk 210, a region proposal network (RPN) 220, detection head 230, keypoint head 240, and segmentation head 250. Each of these components may be configured as a neural network. Conceptually, in the architecture shown, the trunk 210 is configured to process an input image 201 and prepare a feature map (e.g., an inception of convolutional outputs) that represents the image 201. The RPN 220 takes the feature map generated by the trunk 210 and outputs N number of proposed regions of interest (RoIs) that may include objects of interest, such as people, cars, or any other types of objects. The detection head 230 may then detect which of the N RoIs are likely to contain the object(s) of interest and output corresponding object detection indicators 279, which may define a smaller region, such as a bounding box, of the image 201 that contains the object of interest. In particular embodiments, a bounding box may be the smallest or near smallest rectangle (or any other geometric shape(s)) that is able to fully contain the pixels of the object of interest. For the RoIs deemed to be sufficiently likely to contain the object of interest, which may be referred to as target region definitions, the keypoint head 240 may determine their respective keypoint mappings 289 and the segmentation head 250 may determine their respective segmentation masks 299. In particular embodiments, the detection head 230, keypoint head 240, and segmentation head 250 may perform their respective operations in parallel. In other embodiments, the detection head 230, keypoint head 240, and segmentation head 250 may not perform their operations in parallel but instead adopt a multi-staged processing approach, which has the advantage of reducing computation and speeding up the overall operation. For example, the keypoint head 240 and segmentation head 250 may wait for the detection head 230 to identify the target region definitions corresponding to RoIs that are likely to contain the object of interest and only process those regions. Since the N number of RoIs initially proposed by the RPN 220 is typically much larger than the number of RoIs deemed sufficiently likely to contain the object of interest (e.g., on the order of 1000-to-1, 100-to-1, etc., depending on the image given), having such an architectural configuration could drastically reduce computations performed by the keypoint head 240 and segmentation head 250, thereby enabling the operation to be performed on devices that lack sufficient hardware resources (e.g., mobile devices).
At step 320, the system may generate a feature map for the image using a trunk 210. In particular embodiments, the trunk 210 may be considered as the backbone neural network that learns to represent images holistically and is used by various downstream network branches that may be independently optimized for different applications/tasks (e.g., the RPN 220, detection head 230, keypoint head 240, and segmentation head 250). Conceptually, the trunk 210 is shared with each of the downstream components (e.g., RPN 220, detection head 230, etc.), which significantly reduces computational cost and resources needed for running the overall model.
The trunk 210 contains multiple convolutional layers and generates deep feature representations of the input image. In particular embodiments, the trunk 210 may have a compact architecture that is much smaller compared to ResNet and/or other similar architectures. In particular embodiments, the trunk 210 may include four (or fewer) convolution layers 211, 212, 213, 214, three (or fewer) inception modules 215, 217, 218, and one pooling layer (e.g., max or average pooling) 216. In particular embodiments, each of the convolutional layers 211, 212, 213, 214 may use a kernel size of 3×3 or less. In particular, each input image to the trunk 210 may undergo, in order, a first convolution layer 211 (e.g., with 3×3 kernel or patch size, stride size of 2, and padding size of 1), a second convolution layer 212 (e.g., with 3×3 kernel or patch size, stride size of 2, and padding size of 2), a third convolution layer 213 (e.g., with 3×3 kernel or patch size and dimensionality reduction), another convolution layer 214 (e.g., with 3×3 kernel or patch size), a first inception module 215, a max or average pooling layer 216 (e.g., with 3×3 patch size and stride 2), a second inception module 217, and a third inception module 218.
In particular embodiments, each of the inception modules 215, 217, 218 may take the result from its previous layer, perform separate convolution operations on it, and concatenate the resulting convolutions. For example, in one inception module, which may include dimension reduction operations, the result from the previous layer may undergo: (1) a 1×1 convolution, (2) a 1×1 convolution followed by a 3×3 convolution, (3) a 1×1 convolution followed by a 5×5 convolution, and/or (4) a 3×3 max pooling operation followed a 1×1 dimensionality reduction filter. The results of each may then undergo filter concatenation to generate the output of the inception module. In the embodiment described above, the convolutions performed in the inception module use kernel sizes of 5×5 or less; no 7×7 or larger convolution is used in the inception module, which helps reduce the size of the neural net. By limiting the convolution in the inception modules to 5×5 or less, the resulting convolutions and feature maps would be smaller, which in turn means less computation for the subsequent networks (including the networks associated with the downstream components, such as the RPN 220, detection head 230, etc. Although no 7×7 convolution is used in this particular embodiment, 7×7 convolutions may be used in other embodiments.
Referring again to
Referring again to
To address this, particular embodiments, referred to as RoIAlign, removes the harsh quantization of RoIPool by properly aligning the extracted features with the input. This may be accomplished by avoiding any quantization of the RoI boundaries or bins (i.e., use x/16 instead of [x/16]). Particular embodiments may use bilinear interpolation to compute the exact values of the input features at four regularly sampled locations in each RoI bin, and aggregate the result (using max or average). Through RoIAlign, the system may generate a regional feature map of a predefined dimension for each of the RoIs. Particular embodiments may sample four regular locations, in order to evaluate either max or average pooling. In fact, interpolating only a single value at each bin center (without pooling) is nearly as effective. One could also sample more than four locations per bin, which was found to give diminishing returns.
With RoIAlign, the bilinear interpolation used in the feature pooling 225 process is more accurate but requires more computation. In particular embodiments, the bilinear interpolation process may be optimized by precomputing the bilinear-interpolation weights at each position in the grid across batches.
Referring again to
In particular embodiments, the detection head 230 may be configured to process the inception block associated with a given RoI and output a bounding box and a probability that represents a likelihood of the RoI corresponding to the object of interest (e.g., corresponding to a person). In particular embodiments, the inception block may first be processed by average pooling, and the output of which may be used to generate (1) a bounding-box prediction (e.g., using a fully connected layer) that represents a region definition for the detected object (this bounding box coordinates may more precisely define the region in which the object appears), (2) a classification (e.g., using a fully connected layer), and/or (3) a probability or confidence score (e.g., using Softmax function). Based on the classification and/or probability, the detection head 230 may determine which of the RoIs likely correspond to the object of interest. In particular embodiments, all N RoI candidates may be sorted based on the detection classification/probability. The top M RoI, or their respective region definitions (e.g., which may be refined bounding boxes with updated coordinates that better surround the objects of interest), may be selected based on their respective score/probability of containing the objects of interest (e.g., people). The selected M region definitions may be referred to as target region definitions. In other embodiments, the RoI selection process may use non-maximal suppression (NMS) to help the selection process terminate early. Using NMS, candidate RoIs may be selected while they are being sorted, and once the desired M number of RoIs (or their corresponding region definitions) have been selected, the selection process terminates. This process, therefore, may further reduce runtime.
In particular embodiments, once the detection head 230 selects M target region definitions that are likely to correspond to instances of the object of interest (e.g., people), it may pass the corresponding target region definitions (e.g., the refined bounding boxes) to the keypoint head 240 and segmentation head 250 for them to generate keypoint maps 289 and segmentation masks 299, respectively. As previously mentioned, since the M number of region definitions that correspond to people is typically a lot fewer than the N number of initially-proposed RoIs (i.e., M<<N), filtering in this manner prior to having them processed by the keypoint head 240 and segmentation head 250 significantly reduces computation.
In particular embodiments, before processing the M target region definitions using the keypoint head 240 and segmentation head 250, corresponding regional feature maps may be generated (e.g., using RoIAlign) since the M target region definitions may have refined bounding box definitions that differ from the corresponding RoIs. Referring to
Referring to
Particular embodiments may model a keypoint's location as a one-hot mask, and the keypoint head 240 may be tasked with predicting K masks, one for each of K keypoint types (e.g., left shoulder, right elbow, etc.). For each of the K keypoints of an instance, the training target may be a one-hot m×m binary mask in which a single pixel is labeled as a foreground and the rest being labeled as backgrounds (in which case the foreground would correspond to the pixel location of the body part, such as neck joint, corresponding to the keypoint). During training, for each visible ground-truth keypoint, particular embodiments minimize the cross-entropy loss over an m2-way softmax output (which encourages a single point to be detected). In particular embodiments, the K keypoints may still be treated independently. In particular embodiments, the inception block may be input into a deconvolution layer and 2× bilinear upscaling, producing an output resolution of 56×56. In particular embodiments, a relatively high-resolution output (compared to masks) may be required for keypoint-level localization accuracy. In particular embodiments, the keypoint head 240 may output the coordinates of predicted body parts (e.g., shoulders, knees, ankles, head, etc.) along with a confidence score of the prediction. In particular embodiments, the keypoint head 240 may output respective keypoint masks and/or heat maps for the predetermined body parts (e.g., one keypoint mask and/or heat map for the left knee joint, another keypoint mask and/or heat map for the right knee, and so forth). Each heat map may include a matrix of values corresponding to pixels, with each value in the heat map representing a probability or confidence score that the associated pixel is where the associated body part is located.
Referring to
In particular embodiments, a segmentation mask encodes a detected object's spatial layout. Thus, unlike class labels or box offsets that are inevitably collapsed into short output vectors by fully connected (fc) layers, extracting the spatial structure of masks can be addressed naturally by the pixel-to-pixel correspondence provided by convolutions. Particular embodiments may predict an m×m mask from each RoI using a fully convolutional neural network (FCN). This may allow each layer in the segmentation head 250 to maintain the explicit m×m object spatial layout without collapsing it into a vector representation that lacks spatial dimensions. Unlike previous methods that resort to fc layers for mask prediction, particular embodiments may require fewer parameters and may be more accurate. This pixel-to-pixel behavior may require RoI features, which themselves are small feature maps, to be well aligned to faithfully preserve the explicit per-pixel spatial correspondence. The aforementioned feature pooling process termed RoIAlign (e.g., used in the feature pooling layers 225 and 235) may address this need.
Particular embodiments may repeat one or more steps of the process of
At stage 420, a temporary trunk (referenced as Trunktemp in
In particular embodiments, at stage 430, Trunk1 and the various downstream heads (e.g., the detection head, keypoint head, and segmentation head), referred to as Heads1 in
In particular embodiments, each training image sample, during training, may be processed using the temporary Trunktemp and RPNtemp trained in stage 420 to obtain the aforementioned N candidate RoIs. These N RoIs may then be used for training the Trunk1 and the various Heads1. For example, based on the N RoI candidates, the detection head may be trained to select RoI candidates that are likely to contain the object of interest. For each RoI candidate, the machine-learning algorithm may use a bounding-box regressor to process the feature map associated with the RoI and its corresponding ground truth to learn to generate a refined bounding-box that frames the object of interest (e.g., person). The algorithm may also use a classifier (e.g., foreground/background classifier or object-detection classifier for persons or other objects of interest) to process the feature map associated with the RoI and its corresponding ground truth to learn to predict the object's class. In particular embodiments, for the segmentation head, a separate neural network may process the feature map associated with each RoI, generate a segmentation mask (e.g., which may be represented as a matrix or grid with binary values that indicate whether a corresponding pixel belongs to a detected instance of the object or not), compare the generated mask with a ground-truth mask (e.g., indicating the true pixels belonging to the object), and use the computed errors to update the network via backpropagation. In particular embodiments, for the keypoint head, another neural network may process the feature map associated with each RoI, generate a one-hot mask for each keypoint of interest (e.g., for the head, feet, hands, etc.), compare the generated masks with corresponding ground-truth masks (e.g., indicating the true locations of the keypoints of interest), and use the computed errors to update the network via backpropagation. In particular embodiments, the different heads may be trained in parallel.
In particular embodiments, at stage 440, after Trunk1 and the various Heads1 of the machine-learning model have been trained in stage 430, the RPN1 of the model may be trained with Trunk1 being fixed (i.e., the parameters of Trunk1 would remain as they were after stage 430 and unchanged during this training stage). The training dataset at this stage may again include image samples, each having a corresponding ground truth or label, which may include bounding boxes or any other suitable indicators for RoIs appearing in the image sample. Conceptually, this training stage may refine or tailor the RPN1 to propose regions that are particularly suitable for human detection.
At stage 450, once RPN1 has been trained, the various Heads1 (e.g., detection head, keypoint head, and segmentation head) may be retrained with both Trunk1 and RPN1 fixed, in accordance with particular embodiments (i.e., the parameters of Trunk1 and RPN1 would remain as they were after stage 440 and unchanged during this training stage). The training dataset may be similar to the one used in stage 430 (e.g., each training image sample has known ground-truth bounding boxes, keypoints, and segmentation masks). The training process may also be similar to the process described with reference to stage 430, but now Trunk1 would be fixed and the N candidate RoIs would be generated by the trained (and fixed) Trunk1 and RPN1, rather than the temporary Trunktemp and RPNtemp.
Referring back to
Particular embodiments may repeat one or more stages of the training process of
In particular embodiments, at inference time, the trained machine-learning model may be running on a mobile device with limited hardware resources. Compared with mobile CPU, running operations on mobile GPU may provide significant speed-up. Certain mobile platforms may provide third-party GPU processing engines (e.g., Qualcomm® Snapdragon™ Neural Processing Engine (SNPE)) that allow the trained machine-learning model to utilize the various computing capabilities available (e.g., CPU, GPU, DSP). Such third-party processing engines, however, may have certain limitations that would result in suboptimal runtime performance of the machine-learning model.
One issue with third-party processing engines such as SNPE is that it may only support optimized processing for three-dimensional data (hereinafter referred to as 3D tensors). As previously discussed, the RPN is trained to process a given input image and generate N candidate RoIs, and the detection head is trained to select M RoIs. Each of the RoIs may have three-dimensional feature maps (i.e., channel C, height H, and width W). As such, the model needs to process four-dimensional data. Since processing engines such as SNPE only supports three-dimensional convolution processing, one way to process the feature maps is to process the feature maps of the N RoIs (or M, depending on the stage of the inference process) iteratively, such as using a FOR-loop. The FOR-loop for performing sequential convolutions, for instance, may be as follows:
To avoid the aforementioned issues and improve performance, particular embodiments utilize a technique that transforms the 4D tensor (i.e., the feature maps of the N or M RoIs) into a single 3D tensor so that SNPE may be used to perform a single optimized convolution operation on all the feature maps.
Particular manners of tiling may be more efficient than others, depending on how the combined tensor 550 is to be processed. For example, if the operation to be performed is convolution (e.g., in the initial inception module of each of the heads 230, 240, or 250), the feature maps may be tiled in a certain dimension to improve subsequent convolution efficiency (e.g., by improving cache-access efficiency and reducing cache misses). The regional feature map of an RoI can usually be thought of as being three-dimensional, with a height size (H), a width size (W), and a channel size (C). Since the RPN 220 outputs N RoIs, the dimensionality of the N RoIs would be four-dimensional (i.e., H, W, C, and N). The corresponding data representation may be referred to as 4D tensors. In particular embodiments, the 4D tensors may be stored in a data structure that is organized as NCHW (i.e., the data is stored in cache-first order, or in the order of batch, channel, height, and weight). This manner of data storage may provide the detection head with efficient cache access when performing convolution. Similarly, when the segmentation head and/or keypoint head performs convolutions on the regional feature maps of the M region definitions from the detection head, the data may be stored in MCHW order. However, when it comes to the aforementioned feature pooling 225/235 process (e.g., RoIAlign), cache access is more efficient in NHWC or MHWC order, because it can reduce cache miss and utilize SIMD (single instruction multiple data). Thus, in particular embodiments, the feature pooling 225 or 235 process may include a step that organizes or transforms a 4D tensor into NHWC format. This order switching could speed up the feature pooling 225 process significantly.
In particular embodiments, it may be more desirable to generate a large 3D tensor with a larger aspect ratio, such as expanding in only one dimension (i.e., in a row), in order to minimize padding (which in turn minimizes the size of the resulting 3D tensor). Since padding is added between adjacent feature maps, minimizing the surface area of feature maps that are adjacent to other feature maps would result in a reduced need for padding. To illustrate, if N=4, tiling the four feature map tiles in a row may need 3*m padding (i.e., one between the first and second tiles, one between the second and third tiles, and one between the third and fourth tiles). However, if the four feature maps tiles are tiled in a 2×2 configuration, the number of paddings needed would be 4*m (i.e., one between the top-left tile and the top-right tile, one between the top-right tile and the bottom-right tile, one between the bottom-right tile and the bottom-left tile, and one between the bottom-left tile and the top-left tile). Thus, in particular embodiments, additional optimization may be gained by arranging the feature maps in a row (i.e., expanding in one dimension only).
At step 620, the system may generate a feature map that represents the image. In particular embodiments, the system may use the backbone neural network, such as the trunk 210 described herein, to generate the feature map. While other types of backbones (e.g., ResNet, Feature Pyramid Network, etc.) may alternatively be used, embodiments of the trunk 210 provide the advantage of, e.g., not requiring significant hardware resources (e.g., CPU, GPU, cache, memory, etc.) to generate feature maps within stringent timing constraints. Embodiments of the trunk enables applications running on mobile platforms, for example, to take advantage of real-time or near real-time instance detection, classification, segmentation, and/or keypoint generation.
At step 630, the system may identify a regions of interest (RoIs) in the feature map. In particular embodiments, the RoIs may be identified by a region proposal network (RPN), as described herein.
At step 640, the system may generate regional feature maps for the RoIs, respectively. For example, the system may use sampling methods such as RoIPool or RoIAlign to sample an RoI and generate a representative regional feature map. Each of the M regional feature maps generated may have three dimensions (e.g., corresponding to the regional feature map's height, width, and channels). In particular embodiments, the regional feature maps may have equal dimensions (e.g., same height, same width, and same channels).
At step 650, the system may generate a combined regional feature map by combining the M regional feature maps into one. As described previously, the regional feature maps (or 3D tensors) may effectively be tiled together to form a larger 3D tensor. In particular embodiments, the combined regional feature map may be formed by tiling the regional feature maps in a single dimension (e.g., in a row). For example, a first dimension and a second dimension of the combined regional feature map may be equal to the first dimension and the second dimension of each of the plurality of regional feature maps, respectively, with the third dimension of the combined regional feature map being equal to or larger than (e.g., due to added paddings) a combination of the respective third dimensions of the regional feature maps (e.g., the regional feature maps may be stacked in the channel direction, resulting in the combined regional feature map retaining the same height and width as those of an individual regional feature map, but with a larger channel depth). In particular embodiments, the combined regional feature map may be formed by tiling the regional feature maps in two dimensions (e.g., 6 regional feature maps may be tiled in a 2×3 configuration). For example, if the regional feature maps are tiled in both the height and width dimensions, the resulting combined regional feature map's height and width may be larger than a single regional feature map, but its channel would remain the same. In particular embodiments, the combined regional feature map may be formed by tiling the regional feature maps in all three dimensions. In this case, the height, width, and channel of the combined regional feature map may be larger than those of a single regional feature map.
In particular embodiments, to prevent cross sampling from feature maps of different RoIs, the system may insert padding data between adjacent pairs of regional feature maps in the combined regional feature map. In particular embodiments, the size of the padding between each adjacent pair of regional feature maps may be at least as wide as a kernel size used by the one or more convolutional layers.
At step 660, the system may process the combined regional feature map using one or more convolutional layers to generate another combined regional feature map. For example, the system may use a neural processing engine, such as SNPE, to perform the convolutional operations on the combined regional feature map to generate a second combined regional feature map. By combining M regional feature maps into one, the system is able to process the M maps using existing functionalities of neural processing engines, which may be configured to perform convolutional operations only on three-dimensional tensors, thereby taking full advantage of the optimizations offered by such engines.
At step 670, the system may generate, for each of the RoIs, information associated with an object instance based on a portion of the second combined regional feature map (i.e., the result of the convolutional operations performed in step 660) associated with that region of interest. In particular embodiments, for each RoI, the system may identify a region in the second combined regional feature map that corresponds to that RoI. That region may then be used to generate the desired output, such as information associated with an object instance in the RoI. For example, the information associated with the object instance may be an instance segmentation mask, a keypoint mask, a bounding box, or a classification.
Particular embodiments may repeat one or more steps of the process of
In particular embodiments, training of the neural networks for performing convolutional operations on the combined regional feature map may similarly be based on combinations of regional feature maps that are generated during training. For example, for each training image, a training system may generate a feature map (e.g., using the aforementioned trunk), identify RoIs (e.g., using an RPN), generate regional feature maps for the RoIs (e.g., using RoIAlign), and combine the regional feature maps into a larger combined regional feature map, in a similar manner described above. The neural network that is being trained may then process the combined regional feature map to generate a second combined regional feature map. The system may then compare results based on the second combined regional feature map to the corresponding ground truths, and use the comparison results (e.g., as defined by loss functions) to update the neural network. In particular embodiments, the system may, for each RoI, identify a corresponding region in the second combined regional feature map and use it to generate an output (e.g., segmentation mask, bounding box, keypoints, etc.). Each output may then be compared with its corresponding ground truth. In particular embodiments, the ground truths may similarly be tiled to form a combined ground truth. In particular embodiments, the ground truths may have relative positions in the combined ground truth that mirror the relative positions of the corresponding regional feature maps in the combined regional feature map. For example, if the combined regional feature map includes regional feature maps A, B, and C in a row and in that order, the combined ground truth may also include corresponding ground truths A, B, and C in the same order. The error or loss of the predictions may then be computed based on a comparison between the output of the convolutional layers and the combined ground truth.
In particular embodiments, the trained machine-learning model's runtime may further be improved by selectively allocating particular operations to different types of resources. At inference time, different operations may be assigned to different types of processing resources, such as CPU and GPU. Certain segments of operations within the model's operation tree (e.g., a logical representation of the operations performed by the model) may be assigned to a specified type of processing resource that is better suited for performing that type of operation (e.g., more efficient, support the type of data, etc.). Selection of the optimal allocation may be based on benchmarking. For example, different combinations of operation segments may be allocated for processing by a device's GPU or CPU, and the resulting performance and the time needed to transfer the results to another type of processing unit (as well as any other metrics) may be used to determine how best to segment and allocate the operations. In particular embodiments, the benchmarking process may be automatically performed by automatically segmenting the operation tree in different combinations and/or allocating the segments to different computing resources in different combinations. The end performance results may then be ranked to determine which segmentation and/or allocation combination yields the best result.
As described, the machine-learning model according to particular embodiments is compact and optimized for inference-time speed. Such optimizations may, in certain circumstances, result in the accuracy of the prediction results to be less than optimal. To compensate, particular embodiments may perform post-processing to correct or adjust the model's predictions. In particular, the keypoints predictions generated by the keypoint head may be automatically corrected based on a pose model. At a high level, the pose model may learn the poses that humans are likely to make. Using the post model, the keypoints predictions generated by the keypoint head may be automatically adjusted to reflect the more likely poses that the pose model has learned.
In particular embodiments, a two-dimensional (2D) body pose may be represented by a vector S=[x0, y0, x1, y1, . . . , xN-1, yN-1]T that concatenates x and y coordinates of all the keypoints. In particular embodiments, human poses may be represented by any number of keypoints (e.g., N=5, 10, 19, 30, etc.), as described above. Each (x, y) coordinate in S may be defined in an implicit coordinate system of the image. For example, the top-left corner (or any other point) of the image may be defined as the origin (0, 0) of the coordinate space and all other coordinates may be defined relative to the origin. In particular embodiments, each pose S may be normalized, via a transformation function denoted r(S), to a local coordinate system that is defined to be relative to one or more of the predetermined parts of the body represented by the pose. For example, each transformed coordinate (x′, y′) in r(S), which continues to indicate a joint location, may be defined relative to, e.g., the points corresponding to the body's head, shoulders, and/or hips (e.g., the origin (0, 0) may be defined to be the head, the midpoint between the shoulders, the midpoint between the hips, or the midpoint between the respective midpoints of the shoulders and hips). The r(S) coordinates, therefore, may be considered as a set of normalized coordinates. After a local coordinate system for the pose is defined based on its shoulders and hips, for example, a transformation matrix M between the original implicit coordinate system and the local coordinate system may be defined. In particular embodiments, r(S) may be a function that applies M to each (x, y) coordinate in S.
A morphable 2D pose representation of the pose S may be generated based on Principal Component Analysis (PCA), in accordance with particular embodiments. A PCA transformation model may be trained based a training dataset D containing t training poses S1 . . . St, each of which may be a vector of concatenated coordinates of keypoints that correspond to predefined joints or other body parts. The poses in the training dataset D may all be normalized into the same local coordinate system by applying r(S) to each of S1 . . . St. An average of all r(S) in the dataset D may be represented as mean pose Sm:
In particular embodiments, a new pose S(B) may be generated using the PCA model based on the following definition:
S(B)=Sm+Vk*B,
where Vk denotes the first k eigenvectors (e.g., corresponding to the top k principal components with the most variance), Sm denotes mean pose, and B=[β0, β1, . . . , βk-1] denotes the low-dimensional representation of the pose. For a given pose S (or its normalized counterpart r(S)), the pose representation B could be computed as:
B(S)=VkT*(S−Sm)
In particular embodiments, to further constrain the low-dimensional space of pose representation B, a pose-representation probability model for each dimension of B may be learned based on the poses projected to the low-dimensional space. Each dimension may be modeled by a Gaussian distribution gi and the pose-representation probability model of B may be defined as:
A(B)=Πgi(Bi)
In particular embodiments, the pose prediction as represented by the keypoints generated by the keypoint head of the machine-learning model may be adjusted as follows. In particular embodiments, the keypoint head may output a plurality of pose probability models H corresponding to a plurality of body parts (e.g., joints), respectively. The plurality of pose probability models may be configured for determining a probability of the associated body part being at a location in the image. For example, each pose probability model may be a heat map that indicates, for each location represented in the heat map, a probability or confidence score that the associated body part is located at that location. Given the pose probability models H, a post-processing objective may be to find a high-likelihood pose S that best fits the following formulation:
S*=arg minS{−log(Π(Hi(Si)))−log(A(VkT*(r(S)−Sm)))+α*∥S(B(r(S)))−r(S)∥2}
where Si is the i-th coordinate (xi, yi) in S that corresponds to a predetermined i-th body part (e.g., a particular joint), Hi(Si) is the confidence or likelihood of the i-th joint or body part in S being at position (xi, yi) in the image, r(S) is the pose S normalized to the local coordinate, S(B(r(S))) represents the reprojected pose of the pose representation B that lies on the underlying low-dimensional space, and α is the weight between two terms. This problem could be solved by gradient-based optimization or any other suitable optimization technique.
In particular embodiments, to speed up the process, the optimization may be approximated using a discrete approach. For each joint heat-map, the first few (e.g., one or two) local maxima may be found and used as candidates using mean shift algorithm based on the type of joints. For each combination of the candidates, the cost of the pose S may be computed as:
E=−log(Π(Hi(Si)))−log(A(VkT*(r(S)−Sm)))+α*∥S(B(r(S)))−r(S)∥2
The pose with the minimal cost may be used as the final pose.
At step 820, the system may determine a candidate pose that is defined by a set of coordinates representing candidate locations of the predetermined parts of the body in the image. The candidate pose may be determined as part of the optimization process (e.g., either via a gradient-based or discrete-based approach). The candidate pose may be the aforementioned S, defined in the implicit coordinate system.
At step 830, the system may determine a probability score for the candidate pose based on the plurality of pose probability models and the set of coordinates of the candidate pose. For example, the probability score may be based on Hi(Si), where the probability of each keypoint Si of the pose S is looked up in the corresponding pose probability model (e.g., a heat map) Hi to determine the likelihood of that keypoint being correct. The overall probability score may then be computing by, e.g., multiplying the Hi(Si) of every i-th joint in S.
At step 840, the system may generate a pose representation for the candidate pose using a transformation model (e.g., based on PCA) and the candidate pose. In particular embodiments, the transformation model may be applied to the candidate pose S (i.e., B(S)), in which the coordinate of each keypoint is defined in an implicit coordinate system. In other embodiments, the transformation model may be applied to a corresponding set of normalized coordinates defined in a local coordinate (i.e., B(r(S))). In particular embodiments, the transformation model may apply PCA eigenvectors to differences between the set of normalized coordinates (e.g., r(S)) and an aggregate representation of a plurality of sets of normalized coordinates that are associated with a plurality of poses (e.g., Sm), respectively. The resulting pose representation B may be defined in a spatial dimension that is of lower dimensionality than S or r(S).
At step 850, the system may determine a second probability score for the pose representation B based on a pose-representation probability model, such as the model A described above. For example, the system may use a Gaussian distribution model, generated based on known B values for a set of training images, to determine a probability or likelihood of each individual point in B being correct. The aggregate (e.g., based on a multiplication) of the individual probabilities may then represent the probability score for the pose representation B being correct.
In particular embodiments, either in parallel with or sequential to step 850, the system may further compute a reprojection error. For example, at step 860, the system may reproject the pose representation B from its spatial dimension into another spatial dimension associated with the coordinate system of the pose that is represented by the pose representation. For example, if the pose representation B(r(S)) represents the normalized coordinates r(S), then the reprojection S(B(r(S))) may be from the spatial dimension of B back into the spatial dimension of r(S). Similarly, if the pose representation B(S) represents the candidate pose S, then the reprojection S(B(S)) may be from the spatial dimension of B back into the spatial dimension of S. At step 870, the system may then compute a reprojection error based on the reprojected pose representation and the original coordinates represented by B (e.g., S(B(r(S)))−r(S) or S(B(S))−S).
At step 880, the system may determine whether the candidate pose S satisfies one or more criteria, based on one or more of the metrics determined above in steps 830, 850, and 870. As discussed, the criteria may be formulated as an optimization problem, where the objective is to find the S that optimizes the metrics (e.g., maximizes the probability of correctness for the candidate pose, maximizes the probability of correctness for the pose representation, and/or minimizing the reprojection error). In particular embodiments, the previous steps may be repeated iteratively until such a candidate S is determined.
At step 890, the system may select the candidate pose to represent a pose of the body depicted in the image based on at least the first probability score for the candidate pose, the second probability score for the pose representation, and/or the reprojection error.
Particular embodiments may repeat one or more steps of the process of
In particular embodiments, keypoint predictions may be refined using post-processing techniques to improve localization precision. Keypoint predictions, in particular embodiments, may estimate the locations of certain features of interest. In the context of body-pose estimation, for example, the features of interest may be particular joints of interest (e.g., ankles, hip joints, knees, etc.), facial features (e.g., left eye, right ear, head, mouth, etc.), or any other body parts (e.g., head, left hand, right foot, etc.). A keypoint prediction for a particular feature of interest (e.g., left knee or right hip joint) may be represented by a probability model, such as a heat map. The probability model, which may be output by the keypoint head in certain embodiments, may be generated from and correspond to an input image. The probability model may include probability values that are associated with different regions of the input image, with each probability value indicating the probability of the feature of interest being in the associated region. For example, a probability model may be a grid (e.g., 56×56) of probability values that map to a grid of regions in the image. As an example, if the probability value corresponding to the top-left region of the image is 0.67, it may mean that the system predicts a 67% probability or confidence that the body part of interest (e.g., left knee) is located within that region of the image. As such, the larger the size of each region, the less precise the localization is for the feature of interest. For example, if one image is divided into 100×100 regions and another image of the same size is divided into 10×10 regions, each region in the 100×100 regions would be much smaller than each region in the 10×10 regions (the 100×100 division is said to have relatively higher resolution than the 10×10 division). The smaller region can more precisely define the scope of where the feature of interest is located than the larger region.
While the localization precision of a higher-resolution probability model or heat map is higher, the computational resources and time needed to compute all the probability values are also higher. To minimize computation resources and time, especially for computing devices with limited resources (e.g., mobile phones), it may be desirable to limit the resolution of probability models at the expense of precision. Having a low-resolution probability model, however, means that the size of each region would be relatively larger, which in turn means that the estimated location of the joint or feature of interest is less precise.
To improve precision, particular embodiments described herein are directed to systems and methods for up-sampling a given probability model to more precisely pin-point the likely location where the feature of interest is located in the image. In particular embodiments, up-sampling may be applied to the probability values associated with all regions in the image. This may be accomplished by applying bicubic interpolation or convolution/deconvolution to the probability values. Unfortunately, such methods are computationally intensive and may not be suitable for devices with limited processing resources or applications requiring fast outputs (e.g., real-time or near real-time outputs). To address this shortcoming, other embodiments, described in further detail below, provide a much more efficient process for up-sampling a probability model. For example, for a probability model with a 56×56 resolution, the approximate speedup in accordance with such embodiments is roughly 100-times faster than the aforementioned process of applying bicubic interpolation or convolution processes to the entire probability model.
To illustrate,
In particular embodiments, the up-sampling process may compare the probability values in the probability model to identify n number (e.g., 2, 3, 5, etc.) of candidate probability values to further examine. To ensure computational efficiency, the number n may be limited to be less than a fraction (e.g., 1/100, 1/1000, etc.) of the total number of regions (e.g., for 50×50 regions, fewer than one-hundredth, or 25, candidates may be selected).
Each selected candidate probability value may be used to identify surrounding probability values. The regions corresponding to the selected candidate probability value and its surrounding probability values may be referred to as a patch.
Using the identified patches (e.g., patches 1030b, 1030c, and 1030d), the computing system may determine precise locations within the patches that are most likely to be where the feature of interest is located.
In particular embodiments, a final keypoint location for the feature of interest may be selected from the locations associated with the probabilistic maxima of the patches. For example, the maximum probability values computed using quadrilateral interpolation may be compared to find the one that is largest. The final keypoint location for the feature of interest may be set as the location that corresponds to the largest maximum probability value. The coordinates of the location corresponding to the maximum interpolated probability value could be on a continuous spectrum (i.e., not discrete) and therefore could represent a precise location of the feature of interest.
At step 1120, the system may select a subset of the probability values based on a comparison of the probability values. For example, from 56×56 probability values, the system may select the 3 highest values or the 3 highest local maxima. As an example,
At step 1130, the system may identify, for each selected probability value, surrounding probability values whose associated regions surround the region associated with the selected probability value. As an example, in
At step 1140, the system may compute, for each selected probability value, a probabilistic maximum based on the selected probability value and the surrounding probability values. For example, based on the probability values in a patch, quadrilateral interpolation may be used to find the probabilistic maximum corresponding to the maximum interpolated probability value in the patch. The probabilistic maximum may be associated with a location (e.g., a point) within the patch (i.e., the regions associated with the selected probability value and the surrounding probability values). An example of a location associated with a probabilistic maximum is shown in
At step 1150, the system may select, based on the probabilistic maxima, one of the locations associated with the probabilistic maxima to represent a determined location in the image that corresponds to the predetermined body part of the body (e.g., a joint of interest). For example,
Particular embodiments may repeat one or more steps of the process of
Further embodiments described herein are directed to improving the accuracy of keypoint detections used for estimating the body pose of a person depicted in an image. As described elsewhere herein, a machine-learning model may be trained to process an image and output keypoint detections that estimate the locations of predetermined joints of interest (e.g., ankles, hip joints, knees, structural joints, etc.). For each joint of interest, the output may be in the form of a joint probability model (e.g., a heat map) with probability values that correspond to regions (e.g., 56×56 grids) of the image. Each probability value may represent the probability of the associated region containing the joint of interest. However, in a given probability model, there may be several regions with sufficiently high probability values, making it difficult to discern which of those regions is the actual region that contains the joint of interest. This problem may be particularly prevalent in situations where the user's limbs are crossed (e.g., the depicted person's legs may be crossed, resulting in his left ankle appearing on his right side). For example, a probability model generated for detecting the left knee of a person in an image may include a probability value of 0.91 for the actual location of the left knee and a probability value of 0.92 for the person's right knee. If the final keypoint selection for the left knee is set to be the region with the largest probability value, the right knee would be incorrectly selected in this example since 0.92 is greater than 0.91.
To improve the accuracy of joint detection, particular embodiments may select a keypoint for a joint of interest based on a probability model for a connecting segment (e.g., for joints or features of a human, the connecting segments may be referred to as bone segments). For example, the machine-learning model (e.g., via the keypoint head or a separate head) may be trained to output, in addition to probability models for predetermined joints, probability models for predetermined segments that connect the joints in a pose model.
A probability model (e.g., heat map) for a connecting segment, like probability models for joints, may include probability values that are each associated with a region in the image. Each of the probability values may represent the likelihood of the associated region overlapping with the connecting segment of interest. For example, a probability model for the left femur (e.g., bone segment 1266) may include probability values that indicate the likely location of the left femur in the image. Similarly, another probability model may be generated for the right femur (e.g., bone segment 1265), left humerus (e.g., bone segment 1259), or any other predetermined segment connecting two of the predetermined joints.
Referring again to
In particular embodiments, the probability models for connecting segments may be used to refine the estimated position of joints. For example, a probability model for a person's left knee may include two candidate regions (e.g., regions with similarly high probability values), and a probability model for the person's left hip may also have two candidate regions. To choose between the candidate regions, the pose-estimation system may use a probability model for the person's left femur. For example, for each possible pairing between the candidate regions for the left knee and the candidate regions for the left hip, the system may use the probability model for the left femur to compute the probability of the left femur being located between each pair of candidate locations. The resulting probability scores of the different pairs of candidate locations may be compared, and the pair with the highest score may be used as the final estimated locations of the person's left knee and left hip. In a similar manner, an overall optimization problem may be solved to select the entire set of joints (e.g., all 19 joints). This may be achieved, in particular embodiments, by maximizing the entire pose's probability scores computed using the associated bone probability models.
In particular embodiments, for each pair of candidate joints, a probability score may be computed based on the corresponding segment probability model to determine the likelihood of those pair of candidates being the joints of interest. For example, to compute the probability score for the pairing between regions 1310b and 1320b, the regions 1341 connecting those two candidate regions 1310b and 1320b may be identified. Since the joint pairing is between candidates for the right hip and right knee, the probability model for the right femur, which connects the right hip and the right knee in the pose model, may be used to compute the probability score. In particular embodiments, the probability values corresponding to the connecting regions 1341 may be looked up using the probability model for the right femur. For example, the ten connection regions 1341 depicted in
In particular embodiments, the computed probability scores may be used to identify the most likely regions for the joints of interest. For example, the probability scores for the connecting regions 1341, 1342, 1343, and 1344 may be compared, and the candidate pair associated with the highest probability score may be selected to represent the keypoints for the joints of interest (e.g., in this case the right hip and/or the right knee). For example, the probability scores for the connecting regions 1341, 1342, 1343, and 1344 may be 0.95, 0.66, 0.70, 0.85, respectively. The low probability scores for the connecting regions 1342 and 1343 may be attributed to the fact that they include regions between the user's legs with low probability values. Since 0.95 is the highest probability value, the associated regions 1310b and 1320b may be deemed the most likely regions for the right hip and right knee, respectively.
Although in the example above the keypoints are selected based on probability scores for the particular segment type connecting the joints of interest (e.g., the right femur connecting the right hip and right knee), the keypoint selection process, in particular embodiments, may also take a more holistic view and consider the probability scores of other connecting segments and joints. For example, to determine the location of the person's right knee, the selection process may consider the probability scores associated with the person's right tibia/fibula in addition to the right femur (both the right femur and the right tibia/fibula share the right knee as a common joint), as well as any other connected portions of the pose model (e.g., the left femur, spine, right shoulder, etc.). Considering the probability scores of other connecting segments is helpful because the end goal is to predict a pose structure in certain embodiments, rather than a joint location in isolation. Moreover, there may be disagreement between the conclusions drawn from isolated analyses of different types of connecting segments (e.g., based on the probability scores for the right tibia/fibula, the likely location of the right knee may be at region 1320c instead of the aforementioned region 1320b). Thus, in particular embodiments, keypoints for the pose may be determined based on a larger structure that includes multiple different connecting segments and joints. For example, selecting the region for the right knee may be based on not just the probability scores associated with the right femur, but also those of the right tibia/fibula, the segment connecting the right hip and the center hip bone, the spine or segment connecting the center hip bone and the mid-point between the shoulders, etc. For each possible pose (e.g., all 19 keypoints) or portion thereof (e.g., the leg portion includes the right hip joint, right knee, and right ankle) that can be formed using the candidate regions, an aggregated probability score may be computed based on individual probability scores for segments within that pose or portion thereof. For example, in the above example, the probability scores associated with the connecting regions 1341, 1342, 1343, and 1344 are compared to determine which is the most likely location of the right femur. If the right tibia/fibula is also considered, the probability scores for 1341 and 1343 may each be aggregated with the probability score, computed using the probability model for the right tibia/fibula, for the regions under the depicted person's 1301 right knee (e.g., the regions below 1320b, which may be one candidate for the right tibia/fibula). Similarly, the probability scores for 1342 and 1344 may each be aggregated with the probability score, computed using the probability model for the right tibia/fibula, for the regions under the depicted person's 1301 left knee (e.g., the regions below 1320c, which may also be a candidate for the right tibia/fibula). The aggregated probability scores may then be compared to determine which pose configuration is more likely.
At step 1420, the system may access a second probability model associated with the image. The second probability model may include second probability values associated with the regions of the image, with each of the second probability values representing a probability of the associated region of the image containing a predetermined second body part (e.g., the right hip joint). In particular embodiments, the predetermined first body part and the predetermined second body part are directly connected in a body-pose model (e.g., the right knee and the right hip are directly connected in the model shown in
At step 1430, the system may access a third probability model associated with the image. The third probability model may include third probability values associated with the regions of the image, with each of the third probability values representing a probability of the associated region of the image containing a predetermined segment (e.g., right femur bone) connecting the predetermined first body part (e.g., right knee) and the predetermined second body part (e.g., right hip).
At step 1440, the system may select a first region and a second region from the regions of the image based on the first probability model. For example, as illustrated in
At step 1450, the system may select a third region from the regions of the image based on the second probability model. For example, as illustrated in
At step 1460, the system may compute, based on the third probability model, a first probability score for regions connecting the first region and the third region. For example, as illustrated in
At step 1470, the system may compute, based on the third probability model, a second probability score for regions connecting the second region and the third region. For example, as illustrated in
At step 1480, the system may select, based at least one the first probability score and the second probability score, the first region to indicate where the predetermined first body part of the body appears in the image. Referring again to
Particular embodiments may repeat one or more steps of the process of
This disclosure contemplates any suitable network 1510. As an example and not by way of limitation, one or more portions of network 1510 may include an ad hoc network, an intranet, an extranet, a virtual private network (VPN), a local area network (LAN), a wireless LAN (WLAN), a wide area network (WAN), a wireless WAN (WWAN), a metropolitan area network (MAN), a portion of the Internet, a portion of the Public Switched Telephone Network (PSTN), a cellular telephone network, or a combination of two or more of these. Network 1510 may include one or more networks 1510.
Links 1550 may connect client system 1530, social-networking system 1560, and third-party system 1570 to communication network 1510 or to each other. This disclosure contemplates any suitable links 1550. In particular embodiments, one or more links 1550 include one or more wireline (such as for example Digital Subscriber Line (DSL) or Data Over Cable Service Interface Specification (DOCSIS)), wireless (such as for example Wi-Fi or Worldwide Interoperability for Microwave Access (WiMAX)), or optical (such as for example Synchronous Optical Network (SONET) or Synchronous Digital Hierarchy (SDH)) links. In particular embodiments, one or more links 1550 each include an ad hoc network, an intranet, an extranet, a VPN, a LAN, a WLAN, a WAN, a WWAN, a MAN, a portion of the Internet, a portion of the PSTN, a cellular technology-based network, a satellite communications technology-based network, another link 1550, or a combination of two or more such links 1550. Links 1550 need not necessarily be the same throughout network environment 1500. One or more first links 1550 may differ in one or more respects from one or more second links 1550.
In particular embodiments, client system 1530 may be an electronic device including hardware, software, or embedded logic components or a combination of two or more such components and capable of carrying out the appropriate functionalities implemented or supported by client system 1530. As an example and not by way of limitation, a client system 1530 may include a computer system such as a desktop computer, notebook or laptop computer, netbook, a tablet computer, e-book reader, GPS device, camera, personal digital assistant (PDA), handheld electronic device, cellular telephone, smartphone, augmented/virtual reality device, other suitable electronic device, or any suitable combination thereof. This disclosure contemplates any suitable client systems 1530. A client system 1530 may enable a network user at client system 1530 to access network 1510. A client system 1530 may enable its user to communicate with other users at other client systems 1530.
In particular embodiments, client system 1530 may include a web browser 1532, such as MICROSOFT INTERNET EXPLORER, GOOGLE CHROME or MOZILLA FIREFOX, and may have one or more add-ons, plug-ins, or other extensions, such as TOOLBAR or YAHOO TOOLBAR. A user at client system 1530 may enter a Uniform Resource Locator (URL) or other address directing the web browser 1532 to a particular server (such as server 1562, or a server associated with a third-party system 1570), and the web browser 1532 may generate a Hyper Text Transfer Protocol (HTTP) request and communicate the HTTP request to server. The server may accept the HTTP request and communicate to client system 1530 one or more Hyper Text Markup Language (HTML) files responsive to the HTTP request. Client system 1530 may render a webpage based on the HTML files from the server for presentation to the user. This disclosure contemplates any suitable webpage files. As an example and not by way of limitation, webpages may render from HTML files, Extensible Hyper Text Markup Language (XHTML) files, or Extensible Markup Language (XML) files, according to particular needs. Such pages may also execute scripts such as, for example and without limitation, those written in JAVASCRIPT, JAVA, MICROSOFT SILVERLIGHT, combinations of markup language and scripts such as AJAX (Asynchronous JAVASCRIPT and XML), and the like. Herein, reference to a webpage encompasses one or more corresponding webpage files (which a browser may use to render the webpage) and vice versa, where appropriate.
In particular embodiments, social-networking system 1560 may be a network-addressable computing system that can host an online social network. Social-networking system 1560 may generate, store, receive, and send social-networking data, such as, for example, user-profile data, concept-profile data, social-graph information, or other suitable data related to the online social network. Social-networking system 1560 may be accessed by the other components of network environment 1500 either directly or via network 1510. As an example and not by way of limitation, client system 1530 may access social-networking system 1560 using a web browser 1532, or a native application associated with social-networking system 1560 (e.g., a mobile social-networking application, a messaging application, another suitable application, or any combination thereof) either directly or via network 1510. In particular embodiments, social-networking system 1560 may include one or more servers 1562. Each server 1562 may be a unitary server or a distributed server spanning multiple computers or multiple datacenters. Servers 1562 may be of various types, such as, for example and without limitation, web server, news server, mail server, message server, advertising server, file server, application server, exchange server, database server, proxy server, another server suitable for performing functions or processes described herein, or any combination thereof. In particular embodiments, each server 1562 may include hardware, software, or embedded logic components or a combination of two or more such components for carrying out the appropriate functionalities implemented or supported by server 1562. In particular embodiments, social-networking system 1560 may include one or more data stores 1564. Data stores 1564 may be used to store various types of information. In particular embodiments, the information stored in data stores 1564 may be organized according to specific data structures. In particular embodiments, each data store 1564 may be a relational, columnar, correlation, or other suitable database. Although this disclosure describes or illustrates particular types of databases, this disclosure contemplates any suitable types of databases. Particular embodiments may provide interfaces that enable a client system 1530, a social-networking system 1560, or a third-party system 1570 to manage, retrieve, modify, add, or delete, the information stored in data store 1564.
In particular embodiments, social-networking system 1560 may store one or more social graphs in one or more data stores 1564. In particular embodiments, a social graph may include multiple nodes—which may include multiple user nodes (each corresponding to a particular user) or multiple concept nodes (each corresponding to a particular concept)—and multiple edges connecting the nodes. Social-networking system 1560 may provide users of the online social network the ability to communicate and interact with other users. In particular embodiments, users may join the online social network via social-networking system 1560 and then add connections (e.g., relationships) to a number of other users of social-networking system 1560 to whom they want to be connected. Herein, the term “friend” may refer to any other user of social-networking system 1560 with whom a user has formed a connection, association, or relationship via social-networking system 1560.
In particular embodiments, social-networking system 1560 may provide users with the ability to take actions on various types of items or objects, supported by social-networking system 1560. As an example and not by way of limitation, the items and objects may include groups or social networks to which users of social-networking system 1560 may belong, events or calendar entries in which a user might be interested, computer-based applications that a user may use, transactions that allow users to buy or sell items via the service, interactions with advertisements that a user may perform, or other suitable items or objects. A user may interact with anything that is capable of being represented in social-networking system 1560 or by an external system of third-party system 1570, which is separate from social-networking system 1560 and coupled to social-networking system 1560 via a network 1510.
In particular embodiments, social-networking system 1560 may be capable of linking a variety of entities. As an example and not by way of limitation, social-networking system 1560 may enable users to interact with each other as well as receive content from third-party systems 1570 or other entities, or to allow users to interact with these entities through an application programming interfaces (API) or other communication channels.
In particular embodiments, a third-party system 1570 may include one or more types of servers, one or more data stores, one or more interfaces, including but not limited to APIs, one or more web services, one or more content sources, one or more networks, or any other suitable components, e.g., that servers may communicate with. A third-party system 1570 may be operated by a different entity from an entity operating social-networking system 1560. In particular embodiments, however, social-networking system 1560 and third-party systems 1570 may operate in conjunction with each other to provide social-networking services to users of social-networking system 1560 or third-party systems 1570. In this sense, social-networking system 1560 may provide a platform, or backbone, which other systems, such as third-party systems 1570, may use to provide social-networking services and functionality to users across the Internet.
In particular embodiments, a third-party system 1570 may include a third-party content object provider. A third-party content object provider may include one or more sources of content objects, which may be communicated to a client system 1530. As an example and not by way of limitation, content objects may include information regarding things or activities of interest to the user, such as, for example, movie show times, movie reviews, restaurant reviews, restaurant menus, product information and reviews, or other suitable information. As another example and not by way of limitation, content objects may include incentive content objects, such as coupons, discount tickets, gift certificates, or other suitable incentive objects.
In particular embodiments, social-networking system 1560 also includes user-generated content objects, which may enhance a user's interactions with social-networking system 1560. User-generated content may include anything a user can add, upload, send, or “post” to social-networking system 1560. As an example and not by way of limitation, a user communicates posts to social-networking system 1560 from a client system 1530. Posts may include data such as status updates or other textual data, location information, photos, videos, links, music or other similar data or media. Content may also be added to social-networking system 1560 by a third-party through a “communication channel,” such as a newsfeed or stream.
In particular embodiments, social-networking system 1560 may include a variety of servers, sub-systems, programs, modules, logs, and data stores. In particular embodiments, social-networking system 1560 may include one or more of the following: a web server, action logger, API-request server, relevance-and-ranking engine, content-object classifier, notification controller, action log, third-party-content-object-exposure log, inference module, authorization/privacy server, search module, advertisement-targeting module, user-interface module, user-profile store, connection store, third-party content store, or location store. Social-networking system 1560 may also include suitable components such as network interfaces, security mechanisms, load balancers, failover servers, management-and-network-operations consoles, other suitable components, or any suitable combination thereof. In particular embodiments, social-networking system 1560 may include one or more user-profile stores for storing user profiles. A user profile may include, for example, biographic information, demographic information, behavioral information, social information, or other types of descriptive information, such as work experience, educational history, hobbies or preferences, interests, affinities, or location. Interest information may include interests related to one or more categories. Categories may be general or specific. As an example and not by way of limitation, if a user “likes” an article about a brand of shoes the category may be the brand, or the general category of “shoes” or “clothing.” A connection store may be used for storing connection information about users. The connection information may indicate users who have similar or common work experience, group memberships, hobbies, educational history, or are in any way related or share common attributes. The connection information may also include user-defined connections between different users and content (both internal and external). A web server may be used for linking social-networking system 1560 to one or more client systems 1530 or one or more third-party system 1570 via network 1510. The web server may include a mail server or other messaging functionality for receiving and routing messages between social-networking system 1560 and one or more client systems 1530. An API-request server may allow a third-party system 1570 to access information from social-networking system 1560 by calling one or more APIs. An action logger may be used to receive communications from a web server about a user's actions on or off social-networking system 1560. In conjunction with the action log, a third-party-content-object log may be maintained of user exposures to third-party-content objects. A notification controller may provide information regarding content objects to a client system 1530. Information may be pushed to a client system 1530 as notifications, or information may be pulled from client system 1530 responsive to a request received from client system 1530. Authorization servers may be used to enforce one or more privacy settings of the users of social-networking system 1560. A privacy setting of a user determines how particular information associated with a user can be shared. The authorization server may allow users to opt in to or opt out of having their actions logged by social-networking system 1560 or shared with other systems (e.g., third-party system 1570), such as, for example, by setting appropriate privacy settings. Third-party-content-object stores may be used to store content objects received from third parties, such as a third-party system 1570. Location stores may be used for storing location information received from client systems 1530 associated with users. Advertisement-pricing modules may combine social information, the current time, location information, or other suitable information to provide relevant advertisements, in the form of notifications, to a user.
In particular embodiments, a user node 1602 may correspond to a user of social-networking system 1560. As an example and not by way of limitation, a user may be an individual (human user), an entity (e.g., an enterprise, business, or third-party application), or a group (e.g., of individuals or entities) that interacts or communicates with or over social-networking system 1560. In particular embodiments, when a user registers for an account with social-networking system 1560, social-networking system 1560 may create a user node 1602 corresponding to the user, and store the user node 1602 in one or more data stores. Users and user nodes 1602 described herein may, where appropriate, refer to registered users and user nodes 1602 associated with registered users. In addition or as an alternative, users and user nodes 1602 described herein may, where appropriate, refer to users that have not registered with social-networking system 1560. In particular embodiments, a user node 1602 may be associated with information provided by a user or information gathered by various systems, including social-networking system 1560. As an example and not by way of limitation, a user may provide his or her name, profile picture, contact information, birth date, sex, marital status, family status, employment, education background, preferences, interests, or other demographic information. In particular embodiments, a user node 1602 may be associated with one or more data objects corresponding to information associated with a user. In particular embodiments, a user node 1602 may correspond to one or more webpages.
In particular embodiments, a concept node 1604 may correspond to a concept. As an example and not by way of limitation, a concept may correspond to a place (such as, for example, a movie theater, restaurant, landmark, or city); a website (such as, for example, a website associated with social-network system 1560 or a third-party website associated with a web-application server); an entity (such as, for example, a person, business, group, sports team, or celebrity); a resource (such as, for example, an audio file, video file, digital photo, text file, structured document, or application) which may be located within social-networking system 1560 or on an external server, such as a web-application server; real or intellectual property (such as, for example, a sculpture, painting, movie, game, song, idea, photograph, or written work); a game; an activity; an idea or theory; an object in a augmented/virtual reality environment; another suitable concept; or two or more such concepts. A concept node 1604 may be associated with information of a concept provided by a user or information gathered by various systems, including social-networking system 1560. As an example and not by way of limitation, information of a concept may include a name or a title; one or more images (e.g., an image of the cover page of a book); a location (e.g., an address or a geographical location); a website (which may be associated with a URL); contact information (e.g., a phone number or an email address); other suitable concept information; or any suitable combination of such information. In particular embodiments, a concept node 1604 may be associated with one or more data objects corresponding to information associated with concept node 1604. In particular embodiments, a concept node 1604 may correspond to one or more webpages.
In particular embodiments, a node in social graph 1600 may represent or be represented by a webpage (which may be referred to as a “profile page”). Profile pages may be hosted by or accessible to social-networking system 1560. Profile pages may also be hosted on third-party websites associated with a third-party system 1570. As an example and not by way of limitation, a profile page corresponding to a particular external webpage may be the particular external webpage and the profile page may correspond to a particular concept node 1604. Profile pages may be viewable by all or a selected subset of other users. As an example and not by way of limitation, a user node 1602 may have a corresponding user-profile page in which the corresponding user may add content, make declarations, or otherwise express himself or herself. As another example and not by way of limitation, a concept node 1604 may have a corresponding concept-profile page in which one or more users may add content, make declarations, or express themselves, particularly in relation to the concept corresponding to concept node 1604.
In particular embodiments, a concept node 1604 may represent a third-party webpage or resource hosted by a third-party system 1570. The third-party webpage or resource may include, among other elements, content, a selectable or other icon, or other inter-actable object (which may be implemented, for example, in JavaScript, AJAX, or PHP codes) representing an action or activity. As an example and not by way of limitation, a third-party webpage may include a selectable icon such as “like,” “check-in,” “eat,” “recommend,” or another suitable action or activity. A user viewing the third-party webpage may perform an action by selecting one of the icons (e.g., “check-in”), causing a client system 1530 to send to social-networking system 1560 a message indicating the user's action. In response to the message, social-networking system 1560 may create an edge (e.g., a check-in-type edge) between a user node 1602 corresponding to the user and a concept node 1604 corresponding to the third-party webpage or resource and store edge 1606 in one or more data stores.
In particular embodiments, a pair of nodes in social graph 1600 may be connected to each other by one or more edges 1606. An edge 1606 connecting a pair of nodes may represent a relationship between the pair of nodes. In particular embodiments, an edge 1606 may include or represent one or more data objects or attributes corresponding to the relationship between a pair of nodes. As an example and not by way of limitation, a first user may indicate that a second user is a “friend” of the first user. In response to this indication, social-networking system 1560 may send a “friend request” to the second user. If the second user confirms the “friend request,” social-networking system 1560 may create an edge 1606 connecting the first user's user node 1602 to the second user's user node 1602 in social graph 1600 and store edge 1606 as social-graph information in one or more of data stores 1564. In the example of
In particular embodiments, an edge 1606 between a user node 1602 and a concept node 1604 may represent a particular action or activity performed by a user associated with user node 1602 toward a concept associated with a concept node 1604. As an example and not by way of limitation, as illustrated in
In particular embodiments, social-networking system 1560 may create an edge 1606 between a user node 1602 and a concept node 1604 in social graph 1600. As an example and not by way of limitation, a user viewing a concept-profile page (such as, for example, by using a web browser or a special-purpose application hosted by the user's client system 1530) may indicate that he or she likes the concept represented by the concept node 1604 by clicking or selecting a “Like” icon, which may cause the user's client system 1530 to send to social-networking system 1560 a message indicating the user's liking of the concept associated with the concept-profile page. In response to the message, social-networking system 1560 may create an edge 1606 between user node 1602 associated with the user and concept node 1604, as illustrated by “like” edge 1606 between the user and concept node 1604. In particular embodiments, social-networking system 1560 may store an edge 1606 in one or more data stores. In particular embodiments, an edge 1606 may be automatically formed by social-networking system 1560 in response to a particular user action. As an example and not by way of limitation, if a first user uploads a picture, watches a movie, or listens to a song, an edge 1606 may be formed between user node 1602 corresponding to the first user and concept nodes 1604 corresponding to those concepts. Although this disclosure describes forming particular edges 1606 in particular manners, this disclosure contemplates forming any suitable edges 1606 in any suitable manner.
This disclosure contemplates any suitable number of computer systems 1700. This disclosure contemplates computer system 1700 taking any suitable physical form. As example and not by way of limitation, computer system 1700 may be an embedded computer system, a system-on-chip (SOC), a single-board computer system (SBC) (such as, for example, a computer-on-module (COM) or system-on-module (SOM)), a desktop computer system, a laptop or notebook computer system, an interactive kiosk, a mainframe, a mesh of computer systems, a mobile telephone, a personal digital assistant (PDA), a server, a tablet computer system, an augmented/virtual reality device, or a combination of two or more of these. Where appropriate, computer system 1700 may include one or more computer systems 1700; be unitary or distributed; span multiple locations; span multiple machines; span multiple data centers; or reside in a cloud, which may include one or more cloud components in one or more networks. Where appropriate, one or more computer systems 1700 may perform without substantial spatial or temporal limitation one or more steps of one or more methods described or illustrated herein. As an example and not by way of limitation, one or more computer systems 1700 may perform in real time or in batch mode one or more steps of one or more methods described or illustrated herein. One or more computer systems 1700 may perform at different times or at different locations one or more steps of one or more methods described or illustrated herein, where appropriate.
In particular embodiments, computer system 1700 includes a processor 1702, memory 1704, storage 1706, an input/output (I/O) interface 1708, a communication interface 1710, and a bus 1712. Although this disclosure describes and illustrates a particular computer system having a particular number of particular components in a particular arrangement, this disclosure contemplates any suitable computer system having any suitable number of any suitable components in any suitable arrangement.
In particular embodiments, processor 1702 includes hardware for executing instructions, such as those making up a computer program. As an example and not by way of limitation, to execute instructions, processor 1702 may retrieve (or fetch) the instructions from an internal register, an internal cache, memory 1704, or storage 1706; decode and execute them; and then write one or more results to an internal register, an internal cache, memory 1704, or storage 1706. In particular embodiments, processor 1702 may include one or more internal caches for data, instructions, or addresses. This disclosure contemplates processor 1702 including any suitable number of any suitable internal caches, where appropriate. As an example and not by way of limitation, processor 1702 may include one or more instruction caches, one or more data caches, and one or more translation lookaside buffers (TLBs). Instructions in the instruction caches may be copies of instructions in memory 1704 or storage 1706, and the instruction caches may speed up retrieval of those instructions by processor 1702. Data in the data caches may be copies of data in memory 1704 or storage 1706 for instructions executing at processor 1702 to operate on; the results of previous instructions executed at processor 1702 for access by subsequent instructions executing at processor 1702 or for writing to memory 1704 or storage 1706; or other suitable data. The data caches may speed up read or write operations by processor 1702. The TLBs may speed up virtual-address translation for processor 1702. In particular embodiments, processor 1702 may include one or more internal registers for data, instructions, or addresses. This disclosure contemplates processor 1702 including any suitable number of any suitable internal registers, where appropriate. Where appropriate, processor 1702 may include one or more arithmetic logic units (ALUs); be a multi-core processor; or include one or more processors 1702. Although this disclosure describes and illustrates a particular processor, this disclosure contemplates any suitable processor.
In particular embodiments, memory 1704 includes main memory for storing instructions for processor 1702 to execute or data for processor 1702 to operate on. As an example and not by way of limitation, computer system 1700 may load instructions from storage 1706 or another source (such as, for example, another computer system 1700) to memory 1704. Processor 1702 may then load the instructions from memory 1704 to an internal register or internal cache. To execute the instructions, processor 1702 may retrieve the instructions from the internal register or internal cache and decode them. During or after execution of the instructions, processor 1702 may write one or more results (which may be intermediate or final results) to the internal register or internal cache. Processor 1702 may then write one or more of those results to memory 1704. In particular embodiments, processor 1702 executes only instructions in one or more internal registers or internal caches or in memory 1704 (as opposed to storage 1706 or elsewhere) and operates only on data in one or more internal registers or internal caches or in memory 1704 (as opposed to storage 1706 or elsewhere). One or more memory buses (which may each include an address bus and a data bus) may couple processor 1702 to memory 1704. Bus 1712 may include one or more memory buses, as described below. In particular embodiments, one or more memory management units (MMUs) reside between processor 1702 and memory 1704 and facilitate accesses to memory 1704 requested by processor 1702. In particular embodiments, memory 1704 includes random access memory (RAM). This RAM may be volatile memory, where appropriate. Where appropriate, this RAM may be dynamic RAM (DRAM) or static RAM (SRAM). Moreover, where appropriate, this RAM may be single-ported or multi-ported RAM. This disclosure contemplates any suitable RAM. Memory 1704 may include one or more memories 1704, where appropriate. Although this disclosure describes and illustrates particular memory, this disclosure contemplates any suitable memory.
In particular embodiments, storage 1706 includes mass storage for data or instructions. As an example and not by way of limitation, storage 1706 may include a hard disk drive (HDD), a floppy disk drive, flash memory, an optical disc, a magneto-optical disc, magnetic tape, or a Universal Serial Bus (USB) drive or a combination of two or more of these. Storage 1706 may include removable or non-removable (or fixed) media, where appropriate. Storage 1706 may be internal or external to computer system 1700, where appropriate. In particular embodiments, storage 1706 is non-volatile, solid-state memory. In particular embodiments, storage 1706 includes read-only memory (ROM). Where appropriate, this ROM may be mask-programmed ROM, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), electrically alterable ROM (EAROM), or flash memory or a combination of two or more of these. This disclosure contemplates mass storage 1706 taking any suitable physical form. Storage 1706 may include one or more storage control units facilitating communication between processor 1702 and storage 1706, where appropriate. Where appropriate, storage 1706 may include one or more storages 1706. Although this disclosure describes and illustrates particular storage, this disclosure contemplates any suitable storage.
In particular embodiments, I/O interface 1708 includes hardware, software, or both, providing one or more interfaces for communication between computer system 1700 and one or more I/O devices. Computer system 1700 may include one or more of these I/O devices, where appropriate. One or more of these I/O devices may enable communication between a person and computer system 1700. As an example and not by way of limitation, an I/O device may include a keyboard, keypad, microphone, monitor, mouse, printer, scanner, speaker, still camera, stylus, tablet, touch screen, trackball, video camera, another suitable I/O device or a combination of two or more of these. An I/O device may include one or more sensors. This disclosure contemplates any suitable I/O devices and any suitable I/O interfaces 1708 for them. Where appropriate, I/O interface 1708 may include one or more device or software drivers enabling processor 1702 to drive one or more of these I/O devices. I/O interface 1708 may include one or more I/O interfaces 1708, where appropriate. Although this disclosure describes and illustrates a particular I/O interface, this disclosure contemplates any suitable I/O interface.
In particular embodiments, communication interface 1710 includes hardware, software, or both providing one or more interfaces for communication (such as, for example, packet-based communication) between computer system 1700 and one or more other computer systems 1700 or one or more networks. As an example and not by way of limitation, communication interface 1710 may include a network interface controller (NIC) or network adapter for communicating with an Ethernet or other wire-based network or a wireless NIC (WNIC) or wireless adapter for communicating with a wireless network, such as a WI-FI network. This disclosure contemplates any suitable network and any suitable communication interface 1710 for it. As an example and not by way of limitation, computer system 1700 may communicate with an ad hoc network, a personal area network (PAN), a local area network (LAN), a wide area network (WAN), a metropolitan area network (MAN), or one or more portions of the Internet or a combination of two or more of these. One or more portions of one or more of these networks may be wired or wireless. As an example, computer system 1700 may communicate with a wireless PAN (WPAN) (such as, for example, a BLUETOOTH WPAN), a WI-FI network, a WI-MAX network, a cellular telephone network (such as, for example, a Global System for Mobile Communications (GSM) network), or other suitable wireless network or a combination of two or more of these. Computer system 1700 may include any suitable communication interface 1710 for any of these networks, where appropriate. Communication interface 1710 may include one or more communication interfaces 1710, where appropriate. Although this disclosure describes and illustrates a particular communication interface, this disclosure contemplates any suitable communication interface.
In particular embodiments, bus 1712 includes hardware, software, or both coupling components of computer system 1700 to each other. As an example and not by way of limitation, bus 1712 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a front-side bus (FSB), a HYPERTRANSPORT (HT) interconnect, an Industry Standard Architecture (ISA) bus, an INFINIBAND interconnect, a low-pin-count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCIe) bus, a serial advanced technology attachment (SATA) bus, a Video Electronics Standards Association local (VLB) bus, or another suitable bus or a combination of two or more of these. Bus 1712 may include one or more buses 1712, where appropriate. Although this disclosure describes and illustrates a particular bus, this disclosure contemplates any suitable bus or interconnect.
Herein, a computer-readable non-transitory storage medium or media may include one or more semiconductor-based or other integrated circuits (ICs) (such, as for example, field-programmable gate arrays (FPGAs) or application-specific ICs (ASICs)), hard disk drives (HDDs), hybrid hard drives (HHDs), optical discs, optical disc drives (ODDs), magneto-optical discs, magneto-optical drives, floppy diskettes, floppy disk drives (FDDs), magnetic tapes, solid-state drives (SSDs), RAM-drives, SECURE DIGITAL cards or drives, any other suitable computer-readable non-transitory storage media, or any suitable combination of two or more of these, where appropriate. A computer-readable non-transitory storage medium may be volatile, non-volatile, or a combination of volatile and non-volatile, where appropriate.
Herein, “or” is inclusive and not exclusive, unless expressly indicated otherwise or indicated otherwise by context. Therefore, herein, “A or B” means “A, B, or both,” unless expressly indicated otherwise or indicated otherwise by context. Moreover, “and” is both joint and several, unless expressly indicated otherwise or indicated otherwise by context. Therefore, herein, “A and B” means “A and B, jointly or severally,” unless expressly indicated otherwise or indicated otherwise by context.
The scope of this disclosure encompasses all changes, substitutions, variations, alterations, and modifications to the example embodiments described or illustrated herein that a person having ordinary skill in the art would comprehend. The scope of this disclosure is not limited to the example embodiments described or illustrated herein. Moreover, although this disclosure describes and illustrates respective embodiments herein as including particular components, elements, feature, functions, operations, or steps, any of these embodiments may include any combination or permutation of any of the components, elements, features, functions, operations, or steps described or illustrated anywhere herein that a person having ordinary skill in the art would comprehend. Furthermore, reference in the appended claims to an apparatus or system or a component of an apparatus or system being adapted to, arranged to, capable of, configured to, enabled to, operable to, or operative to perform a particular function encompasses that apparatus, system, component, whether or not it or that particular function is activated, turned on, or unlocked, as long as that apparatus, system, or component is so adapted, arranged, capable, configured, enabled, operable, or operative. Additionally, although this disclosure describes or illustrates particular embodiments as providing particular advantages, particular embodiments may provide none, some, or all of these advantages.
This application is a continuation-in-part under 35 U.S.C. § 120 of U.S. patent application Ser. No. 15/972,035, filed 4 May 2018, which claims the benefit under 35 U.S.C. § 119(e) of U.S. Provisional Patent Application No. 62/593,980, filed 3 Dec. 2017. This application is a continuation-in-part under 35 U.S.C. § 120 of U.S. patent application Ser. No. 15/971,930, filed 4 May 2018, which claims the benefit under 35 U.S.C. § 119(e) of U.S. Provisional Patent Application No. 62/593,980, filed 3 Dec. 2017. This application is a continuation-in-part under 35 U.S.C. § 120 of U.S. patent application Ser. No. 15/971,997, filed 4 May 2018, which claims the benefit under 35 U.S.C. § 119(e) of U.S. Provisional Patent Application No. 62/593,980, filed 3 Dec. 2017. The U.S. patent application Ser. No. 15/972,035, U.S. patent application Ser. No. 15/971,930, U.S. patent application Ser. No. 15/971,997, and U.S. Provisional Patent Application No. 62/593,980 are incorporated herein by reference.
| Number | Date | Country | |
|---|---|---|---|
| 62593980 | Dec 2017 | US | |
| 62593980 | Dec 2017 | US | |
| 62593980 | Dec 2017 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 15972035 | May 2018 | US |
| Child | 16236974 | US | |
| Parent | 15971930 | May 2018 | US |
| Child | 15972035 | US | |
| Parent | 15971997 | May 2018 | US |
| Child | 15971930 | US |
