Patent Application
20250142172
The present disclosure concerns a method for enriching audiovisual content. It also relates to a system implementing this method, in particular, a digital platform.
The field of the present disclosure includes digital communications, digital marketing, video and e-commerce.
Steep growth is currently being observed in audiovisual content offerings. More and more of this audiovisual content is associated with brands and commercial offers.
Today, if someone viewing one such audiovisual content wishes to purchase a product or service seen therein, they must be redirected, via a digital communication network, to a sales site on the Web and use a payment tool. Examples include “Click and Buy®” and “Google Ads®.”
The disadvantage of current methods is that they do not give users real-time access to product or service purchasing services.
Furthermore, it is difficult for video content creators to include in their video content the technical means to provide viewers of their video creations with simple, direct access for purchases or access to services.
One of the main aims of the present disclosure is to remedy this drawback by proposing an innovative video content enhancement method that is powerful and easy to use.
This objective is achieved with a method for enriching initial video content, comprising the following steps:
The steps of recognizing and classifying can advantageously implement artificial intelligence techniques.
The step of integrating into the initial video content can be arranged to graphically represent a capsule when viewing the enriched video content.
The enrichment method according to the present disclosure may also comprise a step for creating tactile selection zones in the enriched video content, associated with one or more selected elements to which one or more capsules have been associated, so that a tactile action on one of these selection zones causes a capsule associated with it to be displayed.
The capsule creation step can be arranged to include in this capsule an additional video and/or an online store and/or a searchable document such as a press article.
Embodiments of the present disclosure will be better understood in view of the figures, in which:
iFrame: name given to an HTML tag used in computer language to integrate the content of another HTML page into an HTML page.
API: for Application Programming Interface: application programming interface comprising a set of definitions and protocols that facilitate the creation and integration of application software.
Bounding Box: A rectangle aligned with the axis. This is the simplest type of closed planar shape, represented by two points containing the minimum and maximum coordinates of each axis.
PyTorch: An open-source machine learning framework that accelerates the transition from research prototyping to production deployment.
SAAS: for Software as a Service: an application software solution hosted in the cloud and operated outside the organization or company by a third party, also known as a service provider. The SaaS solution is accessible on demand via an Internet connection.
Confidence score: Classification confidence scores are designed to measure the accuracy of the model when predicting class assignment.
YOLOv5: A family of compound-scaled object detection models, trained on the COCO dataset, with simple features for increasing test time (TTA), model assembly, hyperparameter evolution and export to ONNX, CoreML and TFLite.
With reference to
Capsule content created in this way is retrieved by accessing, for example, in the Cloud (digital cloud) or via digital networks or the Web, marketplaces PM1, . . . Pmn as well as content platforms, notably video content platforms, PC1, . . . PCk 41.
With reference to
This algorithm 100 was created using the Py Torch coding tool commonly used in machine learning. In a first step 101, a YOLOv5 compound-scaled object detection model is implemented on COCO data. From this model, cropped person images using bounding boxes 102 (or B.Boxes) are identified. The next step 103 is a bounding box prediction from mode data 107.
The prediction is then subjected to a confidence score test, a well-known technique in machine learning, in step 104. If the confidence score is less than a predetermined value X, then this bounding box B.Box is ignored (step 108). If, on the other hand, the confidence score is greater than X, then the next step is 105 to identify the class confidence, followed by the step 106 to detect the image containing the desired object.
With reference to
In a first step I, the person consults the video content and locates an element, such as a known actor 2. In a second step II, the person then selects this element 2 by touching the screen 1 with a finger of their hand 3.
In a third step III, this selection causes a capsule icon 4 to be displayed in a corner of the screen 1, for example, the top right-hand corner.
The person then selects (step IV) this capsule icon 4, which causes a window 5 to appear on screen 1, integrated into the video, this window 5 comprising three zones 50, 51, 52 corresponding respectively to an additional video, an online store and a press article, these three resources all being related to the selected element 2.
So, if the user selects (step V) the zone 50, a window 20 for viewing an additional video appears and the user can directly watch this additional video within the enriched video content.
If the user selects the zone 51 (step VI), an online store 21 appears, integrated within the enriched video content.
If the user selects (step VII) the zone 52, a window 22 for reading a news article appears, integrated within the enriched video content.
The method for producing enriched audiovisual content can be implemented on a video creation platform. This method enables video content creators to offer their viewers the opportunity to interact within their creation.
The video player can be available as a SAAS application for technology novices and video professionals alike, and allows for the embedding of brands that a creator would like to see integrated into your audiovisual content.
The video player used in the method according to the present disclosure may have the following features:
The enrichment method is designed to classify the capsules created as a result of the recognition process. Artificial intelligence will classify all recognized elements in a capsule integrated into the top right-hand corner of a video.
It will then deduce three offers for the element:
The texts describing these offers can be translated into several languages simultaneously.
For product recognition, artificial intelligence selects products available on e-commerce platforms worldwide. If the products are not recognized on one of the platforms, the artificial intelligence then sends a “product to be created” message to the product creation department, which will create the product in question: dimensions, colors, materials, etc.
The artificial intelligence is designed to perform the following operations:
The user of the communication equipment can touch the screen at any time. This action causes a capsule to appear on the screen. Beforehand, the user will have validated the conditions of use of the product prediction algorithm.
The video player can be equipped with a biometric fingerprint reader, and offer the customer the option of registering data such as banking information, tastes, height, weight, measurements, identity, address, function, etc.
As soon as the customer validates the element on their screen with their finger, the ordered product is sent directly to their home.
Numerous types of capsule can be envisaged. The following is a non-exhaustive list of capsules that can be created as a result of the method for recognition in video content.
As a non-limiting example, 200 capsules for a 3-minute video, 600 capsules for a 6-minute video and over 1000 capsules for a video longer than 6 minutes may be contemplated.
The user of the enrichment method according to the present disclosure integrates video content in the form of a file or link into the Web platform or SAAS application.
The artificial intelligence then processes the video content over a period of time equal to or less than the duration of the video content, depending on the availability of the queue for this process, referencing all the information in a dashboard assigned to the user.
The platform for implementing the enrichment method according to the present disclosure can be designed to group together a plurality of services made available to creators and referred to under the term “Transmedia Universe”:
Thanks to partnerships with e-commerce, fashion, real estate and other industries, it is possible to promise customers a wide range of products. As for video creators, they now have the opportunity to create short, interesting videos about the brands listed.
The use of artificial intelligence contributes to revenue optimization and increased visibility for brands and designers.
If video recognition is successful, a predicted number of capsules will appear in the capsule bar on the right-hand side of your screen. This capsule bar remains frozen for the duration of viewing. You can touch the screen at any time. This action opens the Capsule functions.
If video recognition is not enabled, capsules are not automatically displayed on screen. The capsule bar is therefore deactivated. However, you can touch the screen at any time. This action brings up the Capsule bar, which becomes active on the screen. You can interact with the Capsule, which animates three options.
The artificial intelligence implemented in the enrichment method according to the present disclosure is designed to:
Two bars can be displayed below the video:
A search bar is also offered to the user of the enrichment method according to the present disclosure, to enable them to optimize the relevance of the image or video content that will be associated with the product or service for which it is desired to encourage purchases or user engagement.
An analytics API is provided to perform a search function, which improves the visibility of queries and determines the keywords for displaying the relevant video or images from the Internet.
It also involves defining query rules for problematic queries, or adjusting search attributes/parameters to solve relevant and systemic problems.
Filters are also provided to establish optimized conversion paths by predefining filters on specific keywords based on the most popular filters for video or image searches.
An API from an online store, video or website (photo, video, text) can be stored in the algorithm.
A link is established between the registered API element and the element duplicated in the algorithm.
The element duplicated by Artificial Intelligence is searched for in its database.
When an e-commerce platform makes available products or services, the iFrames of which cannot be directly integrated within video content, it is possible to integrate the products or services with this type of iFrame in a dedicated store to enable the iFrame to be played in full, without leaving the video (otherwise this could be considered click and buy).
It should be noted that the data integrated and therefore duplicated in the algorithm (e.g., over 500 million products—images, texts, videos) can require huge servers and enormous computing times (recognition, linking, recording). An algorithm can then be devised to link the video's publication date with a product's market release date. This has the effect of removing the quantities of searches by the algorithm for products hosted by eCommerce platforms.
Of course, the present disclosure is not limited to the examples that have just been described, and many other embodiments may be envisaged without departing from the scope of this invention. In particular, the number of recognizable elements and capsules that can be integrated into enriched video content is only limited by the capacity and power of the computer servers used.
| Number | Date | Country | Kind |
|---|---|---|---|
| FR2201014 | Feb 2022 | FR | national |
This application is a national phase entry under 35 U.S.C. § 371 of International Patent Application PCT/EP2023/052610, filed Feb. 2, 2023, designating the United States of America and published as International Patent Publication WO 2023/148296 A1 on Aug. 10, 2023, which claims the benefit under Article 8 of the Patent Cooperation Treaty of French Patent Application Serial No. FR2201014, filed Feb. 4, 2022.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/EP2023/052610 | 2/2/2023 | WO |
