Method for training a driving related object detector

Abstract

A method for driving-related object detection, the method may include receiving an input image by an input of an object detector; and detecting, by an object detector, objects that appear in the input image. The detecting includes searching for (i) a first object having a first size that is within a first size range and belongs to a four wheel vehicle class, (ii) a second object having a second size that is within a second size range and belongs to a subclass out of multiple four wheel vehicle subclasses, (iii) a pedestrian, and (iv) a two wheel vehicle; wherein a maximum of the first size range does not substantially exceed a minimum of the second size range.

Description

BACKGROUND

Object detection is required in various systems and applications.

Autonomous vehicles and advanced driver assistance systems (ADAS) may include object detection units that should detect driving related objects such as vehicles and pedestrians.

These object detection units are trained using a supervised machine learning process.

In a supervised machine learning process test images are fed to an object detector that outputs an estimate of the objects that appear in the test images.

An error unit receives the estimate and calculated an error between the estimate and reference information that indicates which objects appeared in the test images.

The error is fed to the object detector in order to adjust the object detector to the actual objects that appeared in the test images.

The error calculation and the adjustment of the object detector may be performed in an iterative manner—for example per test image.

The reference information may be generated manually or by other error prone processes and this may reduce the accuracy of the error calculation.

Furthermore—the object detector may be adjusted to compensate for insignificant differences between objects. This adjustment may deviate the configuration of the object detector from an ideal configuration and may also consume many computational resources.

There is a growing need to provide a method, computer readable medium and a system that may be able to provide highly accurate object detection at a low cost.

SUMMARY

There may be provided a method for driving-related object detection, the method may include receiving an input image by an input of an object detector; and

detecting, by an object detector, objects that appear in the input image; wherein the detecting may include searching for (i) a first object having a first size that may be within a first size range and belongs to a four wheel vehicle class, (ii) a second object having a second size that may be within a second size range and belongs to a subclass out of multiple four wheel vehicle subclasses, (iii) a pedestrian, and (iv) a two wheel vehicle; wherein a maximum of the first size range does not substantially exceed a minimum of the second size range.

The object detector may be trained to detect (i) four wheel vehicle having a size within the first size range and belongs to a four wheel vehicle class, and at least two out of (a) a car having a size within the second size range, (b) a truck having a size within the second size range, (c) a bus having a size within the second size range, and (d) a van having a size within the second size range.

The receiving may be preceded by training the object detector to detect (i) four wheel vehicle having a size within the first size range and belongs to a four wheel vehicle class, and at least two out of (a) a car having a size within the second size range, (b) a truck having a size within the second size range, (c) a bus having a size within the second size range, and (d) a van having a size within the second size range.

Each one of the four wheel vehicle class, at least some of the multiple four wheel vehicle subclasses, may have a unique set of anchors.

Each one of the four wheel vehicle class, two wheel vehicle and the pedestrian may have a unique set of anchors.

The object detector may include a shallow neural network that may be followed by a region unit.

The the region unit may include a dedicated section for each one out of the four wheel vehicle class, the two wheel vehicle and the pedestrian.

The the shallow neural network may include five by file convolutional filters.

The first size range and the second size range may not overlap.

The first size range and the second size range partially overlap.

There may be provided a non-transitory computer readable medium for driving-related object detection by an object detector, the non-transitory computer readable medium stores instructions for receiving an input image by an input of the object detector; and detecting, by an object detector, objects that appear in the input image; wherein the detecting may include searching for (i) a first object having a first size that may be within a first size range and belongs to a four wheel vehicle class, (ii) a second object having a second size that may be within a second size range and belongs to a subclass out of multiple four wheel vehicle subclasses, (iii) a pedestrian, and (iv) a two wheel vehicle; wherein a maximum of the first size range does not substantially exceed a minimum of the second size range.

The object detector may be trained to detect (i) four wheel vehicle having a size within the first size range and belongs to a four wheel vehicle class, and at least two out of (a) a car having a size within the second size range, (b) a truck having a size within the second size range, (c) a bus having a size within the second size range, and (d) a van having a size within the second size range.

The non-transitory computer readable medium that stores instructions for training the object detector to detect (i) four wheel vehicle having a size within the first size range and belongs to a four wheel vehicle class, and at least two out of (a) a car having a size within the second size range, (b) a truck having a size within the second size range, (c) a bus having a size within the second size range, and (d) a van having a size within the second size range.

Each one of the four wheel vehicle class, at least some of the multiple four wheel vehicle subclasses, may have a unique set of anchors.

Each one of the four wheel vehicle class, two wheel vehicle and the pedestrian may have a unique set of anchors.

The object detector may include a shallow neural network that may be followed by a region unit.

The non-transitory computer readable medium wherein the region unit may include a dedicated section for each one out of the four wheel vehicle class, the two wheel vehicle and the pedestrian.

The non-transitory computer readable medium wherein the shallow neural network may include five by file convolutional filters.

The first size range and the second size range may not overlap.

The first size range and the second size range partially overlap.

There may be provided an object detector that may include an input, a shallow neural network and a region unit; wherein the region unit follows the shallow neural network; wherein the input may be configured to receive an input image; and wherein the shallow neural network and the region unit may be configured to cooperate and detect objects that appear in the input image; wherein a detecting of the object may include may include searching for (i) a first object having a first size that may be within a first size range and belongs to a four wheel vehicle class, (ii) a second object having a second size that may be within a second size range and belongs to a subclass out of multiple four wheel vehicle subclasses, (iii) a pedestrian, and (iv) a two wheel vehicle; wherein a maximum of the first size range does not substantially exceed a minimum of the second size range.

The object detector may be trained to detect (i) four wheel vehicle having a size within the first size range and belongs to a four wheel vehicle class, and at least two out of (a) a car having a size within the second size range, (b) a truck having a size within the second size range, (c) a bus having a size within the second size range, and (d) a van having a size within the second size range.

The receiving may be preceded by training the object detector to detect (i) four wheel vehicle having a size within the first size range and belongs to a four wheel vehicle class, and at least two out of (a) a car having a size within the second size range, (b) a truck having a size within the second size range, (c) a bus having a size within the second size range, and (d) a van having a size within the second size range.

Each one of the four wheel vehicle class, at least some of the multiple four wheel vehicle subclasses, may have a unique set of anchors.

Each one of the four wheel vehicle class, two wheel vehicle and the pedestrian may have a unique set of anchors.

The region unit may include a dedicated section for each one out of the four wheel vehicle class, the two wheel vehicle and the pedestrian.

The shallow neural network may include five by file convolutional filters.

The first size range and the second size range may not overlap.

The first size range and the second size range partially overlap.

BRIEF DESCRIPTION OF THE DRAWINGS

The embodiments of the disclosure will be understood and appreciated more fully from the following detailed description, taken in conjunction with the drawings in which:

FIG. 1 illustrates an example of a method for object detection;

FIG. 2 illustrates an example of an image;

FIG. 3 illustrates an example of a classification of some of the objects that appear in the image;

FIG. 4 illustrates an example of the image of FIG. 1 with bounding boxes that surround some of the objects that appear in the image;

FIG. 5 illustrates an example of an object detector;

FIG. 6 illustrates an example of various objects; and

FIG. 7 illustrates an example of a training process.

DESCRIPTION OF EXAMPLE EMBODIMENTS

In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be understood by those skilled in the art that the present invention may be practiced without these specific details. In other instances, well-known methods, procedures, and components have not been described in detail so as not to obscure the present invention.

The subject matter regarded as the invention is particularly pointed out and distinctly claimed in the concluding portion of the specification. The invention, however, both as to organization and method of operation, together with objects, features, and advantages thereof, may best be understood by reference to the following detailed description when read with the accompanying drawings.

It will be appreciated that for simplicity and clarity of illustration, elements shown in the figures have not necessarily been drawn to scale. For example, the dimensions of some of the elements may be exaggerated relative to other elements for clarity. Further, where considered appropriate, reference numerals may be repeated among the figures to indicate corresponding or analogous elements.

Because the illustrated embodiments of the present invention may for the most part, be implemented using electronic components and circuits known to those skilled in the art, details will not be explained in any greater extent than that considered necessary as illustrated above, for the understanding and appreciation of the underlying concepts of the present invention and in order not to obfuscate or distract from the teachings of the present invention.

Any reference in the specification to a method should be applied mutatis mutandis to a device or system capable of executing the method and/or to a non-transitory computer readable medium that stores instructions for executing the method.

Any reference in the specification to a system or device should be applied mutatis mutandis to a method that may be executed by the system, and/or may be applied mutatis mutandis to non-transitory computer readable medium that stores instructions executable by the system.

Any reference in the specification to a non-transitory computer readable medium should be applied mutatis mutandis to a device or system capable of executing instructions stored in the non-transitory computer readable medium and/or may be applied mutatis mutandis to a method for executing the instructions.

Any combination of any module or unit listed in any of the figures, any part of the specification and/or any claims may be provided.

There may be provided a low power object detection system (detector), non-transitory computer readable medium, and method. The object detection system, non-transitory computer readable medium and method may prevent errors in learning processes resulting of inaccurate tagging of relatively small four wheel vehicles by classifying (during the training process) objects that are relatively small (appear small in an input image)—within a first size range to a general class of four wheel vehicles.

Larger four wheel vehicle may be classified more accurately to subclasses such as car, bur, truck, van and the like. Other classes (used for at least classifying small objects) include, for example two wheel vehicles and pedestrians.

Each of these other classes may also include subclasses that may be applied to objects that are larger and can be accurately tagged (during the training process) to finer subclasses.

This size based classification increases the accuracy of the reference information received during the training, improves the accuracy of the training process, improves the accuracy of the object detection and also prevents the training process to spend too much resources on attempting to differentiate between insignificant differences. This also provides an object detector that is fine tunes to differentiate between difference that are significant.

After the completion of the training process the object detector it set to detect in an accurate manner the objects according to their size and classes or subclasses.

FIG. 1 illustrates method 9270 for driving-related object detection.

Method 9270 may include the steps of:

• • Step 9272 of receiving an input image by an input of an object detector.
  • Step 9274 of detecting, by an object detector, objects that appear in the input image.

Step 9272 may include searching for (i) a first object having a first size that is within a first size range and belongs to a four wheel vehicle class, (ii) a second object having a second size that is within a second size range and belongs to a subclass out of multiple four wheel vehicle subclasses, (iii) a pedestrian, and (iv) a two wheel vehicle; wherein a maximum of the first size range does not substantially exceed a minimum of the second size range.

The object detector may be trained to perform said search (and method 9270 may include the training—step 9271) by feeding the object detector with images of objects that are tagged according to the searched objects (i) till (iv). The object detector may be also trained to reject objects that are too big and. The training includes feeding the object detector with images that include objects of at least one of the four types (i)-(iv), generating reference information that tags the objects according the the searched types, calculating an error between the outcome of the object detector and the reference information and feeding the error to the object detector.

The subclasses of the four wheel vehicle class may include at least some of the following car, truck, bus, van and the like. The subclasses may include or may be further partitioned to certain type of vehicle, model, and the like.

Accordingly—the classification system may include two layers (single class and single layer of subclasses) or even more layer—to provide a hierarchical classification system that may include more than two layers. For example—a first layer includes a four wheel vehicle, the second layer (subclasses) include car, bus, truck, van and the third layer may include a manufacturer, yet a fourth layer may include model, and the like.

The class of pedestrians and/or the class of two wheel vehicles can also include subclasses to include two or more layers.

Each one of the four wheel vehicle class, at least some of the multiple four wheel vehicle subclasses, may have has a unique set of anchors. The unique sets of anchors may be selected based on the expected shape of the objects that belong to the class and/or the subclass.

Each one of the four wheel vehicle class, two wheel vehicle and the pedestrian has a unique set of anchors.

Anchors may be regarded as initial templates of bounding boxes and using different anchors to different classes may reduce the computational resources, speed the detection and increase the accuracy of the bounding boxes when the different anchors are selected according to the expected shapes of the different vehicles. For example, in a side view, a bus appears longer and higher than a private car. Yet for another example a pedestrian may require bounding boxes that have a height that exceeds their width—while trucks (for example in side view) may require bounding boxes that have a height that is smaller than their width.

FIG. 2 is an example of an image 9301 that includes first bus 9311, second bus 9314, third bus 9316, first truck 9313, second truck 9317, third truck 9318, first car 9315, first bicycle 9319, and first foot scooter 9312.

FIG. 3 illustrates the tagging of the objects of image 9301. The tagging may be included in a reference information used to train the object detector or may be the outcome of the object detection process.

Third bus 9316, second truck 9317, third truck 9318, first car 9315 and first bicycle 9319 are too small (for example have a size within a first size range) and therefore are tagged to classes and not to subclasses.

Third bus 9316, second truck 9317, third truck 9318, first car 9315 are regarded as belonging to a four wheel vehicles class 9321.

The first bicycle 9319 is tagged to belonging to a two wheel vehicle class 9322.

First bus 9311 is large enough (has a size within a second size range) to be tagged as a truck 9325. First bus 9311 and second bus 9314 are large enough (has a size within a second size range) to be tagged as a bus. First foot scooter 9312 is large enough (has a size within a second size range) to be tagged as a foot scooter 9325.

FIG. 4 illustrates an example of a possible output of the object detector—bounding boxes 9310, 9311, 9312, 9313, 9314, 9315, 9316, 9317, 9318 and 9319 that surround pedestrian 9310, first bus 9311, foot scooter 9312, first truck 9313, second bus 9314, first car 9315, third bus 9316, second truck 9317, third truck 9318, first bicycle 9319, respectively.

Each bounding is represented by information 9025 that may include coordinates (x,y,h,w) of the bounding boxes, objectiveness and class. The coordinate indicate the location (x,y) as well as the height and width of the bounding boxes. Objectiveness provides a confidence level that an object exists. Class—class of object—as illustrated above). The (x,y) coordinates may represent the center of the bounding box.

The object detection may be compliant to any flavor of YOLO—but other object detection schemes may be applied.

FIG. 5 illustrates an object detector 9000″ and FIG. 7 illustrates a training process of the object detector.

Object detector 9000″ may include an input 9250, a shallow neural network 9252 and a region unit 9254.

The region unit 9254 follows the shallow neural network 9252.

Input 9250 may be configured to receive an input image 9001.

The shallow neural network 9252 and the region unit 9254 may be configured to cooperate (both participate in the object detection process—the region unit processed the output of the shallow neural network) and detect objects that appear in the input image.

The detecting of the object may include may include searching for (i) a first object having a first size that may be within a first size range and belongs to a four wheel vehicle class, (ii) a second object having a second size that may be within a second size range and belongs to a subclass out of multiple four wheel vehicle subclasses, (iii) a pedestrian, and (iv) a two wheel vehicle; wherein a maximum of the first size range does not substantially exceed a minimum of the second size range.

The shallow neural network 9252 may include convolutional and spooling layers.

The convolutional layers may include convolutional filters. The convolutional filters may be of any shape and size—for example be five by five convolutional filters (have a kernel of five by five elements), be a three by three convolutional filters (have a kernel of three by three elements), and the like.

Object detector 9000″ may be configured to execute method 9270.

The region unit may include a dedicated section for each one out of the four wheel vehicle class, the two wheel vehicle and the pedestrian. This will improve the detection per each class of objects.

At least a part of the object detector may be a processing circuitry that may be implemented as a central processing unit (CPU), and/or one or more other integrated circuits such as application-specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), full-custom integrated circuits, etc., or a combination of such integrated circuits.

At least a part of the object detector may be application implemented in hardware, firmware, or software that may be executed by a processing circuitry.

Regarding the training process:

Test images 9001 are fed to shallow neural network 9252 that outputs, for each test image, a shallow neural network output that may be a tensor with multiple features per segment of the test image. The region unit 9054 is configured to receive the output from shallow neural network 9252 and calculate and output candidate bounding boxes per test image. Actual results such as the output candidate bounding boxes per test image may be fed to error calculation unit 9050.

Error calculation unit 9050 also receives desired results 9045—objects that are tagged to belong to (i) a four wheel vehicle class (if the size with within the first size range), or to (ii) a subclass out of multiple four wheel vehicle subclasses ((if the size with within the first size range), (iii) a pedestrian, and (iv) a two wheel vehicle.

Error calculation unit 9050 calculates an error 9055 between the the actual results and the desired results- and the error is fed to the shallow neural network 9252 during the training process.

It should be noted that in addition to the mentioned above training—the shallow neural network 9252 may be trained to reject (not detect) objects that are too big—for example outside the first and second size ranges.

This may require to train the shallow neural network 9252 not to detect said objects.

FIG. 7 illustrates various objects such as 9031, 9032, 9033 and 9034.

Objects 9033 and 9034 are too big and should be ignored of.

While the foregoing written description of the invention enables one of ordinary skill to make and use what is considered presently to be the best mode thereof, those of ordinary skill will understand and appreciate the existence of variations, combinations, and equivalents of the specific embodiment, method, and examples herein. The invention should therefore not be limited by the above described embodiment, method, and examples, but by all embodiments and methods within the scope and spirit of the invention as claimed.

In the foregoing specification, the invention has been described with reference to specific examples of embodiments of the invention. It will, however, be evident that various modifications and changes may be made therein without departing from the broader spirit and scope of the invention as set forth in the appended claims.

Moreover, the terms “front,” “back,” “top,” “bottom,” “over,” “under” and the like in the description and in the claims, if any, are used for descriptive purposes and not necessarily for describing permanent relative positions. It is understood that the terms so used are interchangeable under appropriate circumstances such that the embodiments of the invention described herein are, for example, capable of operation in other orientations than those illustrated or otherwise described herein.

Furthermore, the terms “assert” or “set” and “negate” (or “deassert” or “clear”) are used herein when referring to the rendering of a signal, status bit, or similar apparatus into its logically true or logically false state, respectively. If the logically true state is a logic level one, the logically false state is a logic level zero. And if the logically true state is a logic level zero, the logically false state is a logic level one.

Those skilled in the art will recognize that the boundaries between logic blocks are merely illustrative and that alternative embodiments may merge logic blocks or circuit elements or impose an alternate decomposition of functionality upon various logic blocks or circuit elements. Thus, it is to be understood that the architectures depicted herein are merely exemplary, and that in fact many other architectures may be implemented which achieve the same functionality.

Any arrangement of components to achieve the same functionality is effectively “associated” such that the desired functionality is achieved. Hence, any two components herein combined to achieve a particular functionality may be seen as “associated with” each other such that the desired functionality is achieved, irrespective of architectures or intermedial components. Likewise, any two components so associated can also be viewed as being “operably connected,” or “operably coupled,” to each other to achieve the desired functionality.

Furthermore, those skilled in the art will recognize that boundaries between the above described operations merely illustrative. The multiple operations may be combined into a single operation, a single operation may be distributed in additional operations and operations may be executed at least partially overlapping in time. Moreover, alternative embodiments may include multiple instances of a particular operation, and the order of operations may be altered in various other embodiments.

Also for example, in one embodiment, the illustrated examples may be implemented as circuitry located on a single integrated circuit or within a same device. Alternatively, the examples may be implemented as any number of separate integrated circuits or separate devices interconnected with each other in a suitable manner.

However, other modifications, variations and alternatives are also possible. The specifications and drawings are, accordingly, to be regarded in an illustrative rather than in a restrictive sense.

In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word ‘comprising’ does not exclude the presence of other elements or steps then those listed in a claim. Furthermore, the terms “a” or “an,” as used herein, are defined as one or more than one. Also, the use of introductory phrases such as “at least one” and “one or more” in the claims should not be construed to imply that the introduction of another claim element by the indefinite articles “a” or “an” limits any particular claim containing such introduced claim element to inventions containing only one such element, even when the same claim includes the introductory phrases “one or more” or “at least one” and indefinite articles such as “a” or “an.” The same holds true for the use of definite articles. Unless stated otherwise, terms such as “first” and “second” are used to arbitrarily distinguish between the elements such terms describe. Thus, these terms are not necessarily intended to indicate temporal or other prioritization of such elements. The mere fact that certain measures are recited in mutually different claims does not indicate that a combination of these measures cannot be used to advantage.

While certain features of the invention have been illustrated and described herein, many modifications, substitutions, changes, and equivalents will now occur to those of ordinary skill in the art. It is, therefore, to be understood that the appended claims are intended to cover all such modifications and changes as fall within the true spirit of the invention.

It is appreciated that various features of the embodiments of the disclosure which are, for clarity, described in the contexts of separate embodiments may also be provided in combination in a single embodiment. Conversely, various features of the embodiments of the disclosure which are, for brevity, described in the context of a single embodiment may also be provided separately or in any suitable sub-combination.

It will be appreciated by persons skilled in the art that the embodiments of the disclosure are not limited by what has been particularly shown and described hereinabove. Rather the scope of the embodiments of the disclosure is defined by the appended claims and equivalents thereof.

Claims

1. A method for driving-related object detection, the method comprises: receiving an input image by an input of an object detector; anddetecting, by the object detector, objects that appear in the input image;wherein the detecting comprises searching for (i) a first object that appears in the image at a first size that is within a first size range and belongs to a four wheel vehicle class, (ii) a second object that appears in the image at a second size that is within a second size range and belongs to a subclass out of multiple four wheel vehicle subclasses, (iii) a pedestrian, and (iv) a two wheel vehicle; wherein a maximum of the first size range is smaller than a minimum of the second size range;wherein the searching for the first object comprises preventing from classifying a four wheel vehicle to any of the multiple four wheel vehicle subclasses when the four wheel vehicle appears in the image at a size that is within the first size range.

2. The method according to claim 1 wherein the object detector is trained to detect (i) four wheel vehicle that appears in the image at a size within the first size range and belongs to a four wheel vehicle class, and at least two out of (a) a car that appears in the image at a size within the second size range, (b) a truck that appears in the image at a size within the second size range, (c) a bus that appears in the image at a size within the second size range, and (d) a van that appears in the image at a size within the second size range.

3. The method according to claim 1 wherein the receiving is preceded by training the object detector to detect (i) four wheel vehicle that appears in the image at a size within the first size range and belongs to a four wheel vehicle class, and at least two out of (a) a car that appears in the image at a size within the second size range, (b) a truck that appears in the image at a size within the second size range, (c) a bus that appears in the image at a size within the second size range, and (d) a van that appears in the image at a size within the second size range.

4. The method according to claim 1 wherein each one of the four wheel vehicle class, at least some of the multiple four wheel vehicle subclasses, has a unique set of anchors.

5. The method according to claim 1 wherein the detecting comprises ignoring objects that appear in the image at a size that exceeds a maximal value of the second size range.

6. The method according to claim 1 wherein the object detector comprises a shallow neural network that is followed by a region unit.

7. The method according to claim 6 wherein the region unit comprises a dedicated section for each one out of the four wheel vehicle class, the two wheel vehicle and the pedestrian.

8. The method according to claim 6 wherein the shallow neural network comprises five by five convolutional filters.

9. The method according to claim 1 wherein each one of the four wheel vehicle class, two wheel vehicle and the pedestrian has a unique set of anchors, the anchors are initial templates of bounding boxes.

10. The method according to claim 9 wherein different anchors are selected according to expected shapes of the vehicles.

11. A non-transitory computer readable medium for driving-related object detection by an object detector, the non-transitory computer readable medium stores instructions for: receiving an input image by an input of the object detector; anddetecting, by the object detector, objects that appear in the input image;wherein the detecting comprises searching for (i) a first object that appears in the image at a first size that is within a first size range and belongs to a four wheel vehicle class, (ii) a second object that appears in the image at a second size that is within a second size range and belongs to a subclass out of multiple four wheel vehicle subclasses, (iii) a pedestrian, and (iv) a two wheel vehicle; wherein a maximum of the first size range is smaller than a minimum of the second size range;wherein the searching for the first object comprises preventing from classifying a four wheel vehicle to any of the multiple four wheel vehicle subclasses when the four wheel vehicle appears in the image at a size that is within the first size range.

12. The non-transitory computer readable medium according to claim 11 wherein the object detector is trained to detect (i) four wheel vehicle that appears in the image at a size within the first size range and belongs to a four wheel vehicle class, and at least two out of (a) a car that appears in the image at a size within the second size range, (b) a truck that appears in the image at a size within the second size range, (c) a bus that appears in the image at a size within the second size range, and (d) a van that appears in the image at a size within the second size range.

13. The non-transitory computer readable medium according to claim 11 that stores instructions for training the object detector to detect (i) four wheel vehicle that appears in the image at a size within the first size range and belongs to a four wheel vehicle class, and at least two out of (a) a car that appears in the image at a size within the second size range, (b) a truck that appears in the image at a size within the second size range, (c) a bus that appears in the image at a size within the second size range, and (d) a van that appears in the image at a size within the second size range.

14. The non-transitory computer readable medium according to claim 11 wherein each one of the four wheel vehicle class, at least some of the multiple four wheel vehicle subclasses, has a unique set of anchors.

15. The non-transitory computer readable medium according to claim 11 wherein the detecting comprises ignoring objects that appear in the image at a size that exceeds a maximal value of the second size range.

16. The non-transitory computer readable medium according to claim 11 wherein the object detector comprises a shallow neural network that is followed by a region unit.

17. The non-transitory computer readable medium according to claim 16 wherein the region unit comprises a dedicated section for each one out of the four wheel vehicle class, the two wheel vehicle and the pedestrian.

18. The non-transitory computer readable medium according to claim 16 wherein the shallow neural network comprises five by five convolutional filters.

19. The non-transitory computer readable medium according to claim 11 wherein each one of the four wheel vehicle class, two wheel vehicle and the pedestrian has a unique set of anchors, the anchors are initial templates of bounding boxes.

20. The non-transitory computer readable medium according to claim 19 wherein different anchors are selected according to expected shapes of the vehicles.

21. An object detector that comprises an input, a shallow neural network and a region unit; wherein the region unit follows the shallow neural network; wherein the input is configured to receive an input image; and wherein the shallow neural network and the region unit are configured to cooperate and detect objects that appear in the input image; wherein a detecting of the object comprises searching for (i) a first object that appears in the image at a first size that is within a first size range and belongs to a four wheel vehicle class, (ii) a second object that appears in the image at a second size that is within a second size range and belongs to a subclass out of multiple four wheel vehicle subclasses, (iii) a pedestrian, and (iv) a two wheel vehicle; wherein a maximum of the first size range is smaller than a minimum of the second size range; wherein the searching for the first object comprises preventing from classifying a four wheel vehicle to any of the multiple four wheel vehicle subclasses when the four wheel vehicle appears in the image at a size that is within the first size range.

22. The object detector according to claim 21 wherein the object detector is trained to detect (i) four wheel vehicle that appears in the image at a size within the first size range and belongs to a four wheel vehicle class, and at least two out of (a) a car that appears in the image at a size within the second size range, (b) a truck that appears in the image at a size within the second size range, (c) a bus that appears in the image at a size within the second size range, and (d) a van that appears in the image at a size within the second size range.

23. The object detector according to claim 21 wherein the receiving is preceded by training the object detector to detect (i) four wheel vehicle that appears in the image at a size within the first size range and belongs to a four wheel vehicle class, and at least two out of (a) a car that appears in the image at a size within the second size range, (b) a truck that appears in the image at a size within the second size range, (c) a bus that appears in the image at a size within the second size range, and (d) a van that appears in the image at a size within the second size range.

24. The object detector according to claim 21 wherein each one of the four wheel vehicle class, at least some of the multiple four wheel vehicle subclasses, has a unique set of anchors.

25. The object detector according to claim 21 wherein the detecting comprises ignoring objects that appear in the image at a size that exceeds a maximal value of the second size range.

26. The object detector according to claim 21 wherein the region unit comprises a dedicated section for each one out of the four wheel vehicle class, the two wheel vehicle and the pedestrian.

27. The object detector according to claim 21 wherein the shallow neural network comprises five by five convolutional filters.

28. The object detector according to claim 21 wherein each one of the four wheel vehicle class, two wheel vehicle and the pedestrian has a unique set of anchors, the anchors are initial templates of bounding boxes.

29. The object detector according to claim 21 wherein different anchors are selected according to expected shapes of the vehicles.