Artificial intelligence is one of the biggest advancements in the technological world. Although machine learning has so far been limited to reading fixed frames and is very effective, the next step in the development of automatic learning and artificial intelligence is recognition of individual objects in the video to provide new possibilities for branding, visual effects, accessibility, etc.
Google has taken the first successful step towards AI recognizing a single object in a video. After a period of hard work, google had introduced new enhancements to the YouTube experience, including the ability to tag products contained in video clips and provide links to directly purchase those products.
This simply means that the company can now tag its products on YouTube videos whenever it is watched. You can mark the product at that specific time. In addition, it provides direct shopping and promotes a wider range of e-commerce opportunities in the market. After successfully implementing this feature on YouTube, Facebook took the next step and introduced similar features on its platform. The company claims that this feature can better identify individual objects in video frames.
Facebook described that they worked with Inria researchers and they had developed a new method called DINO to train the unattended vision transformer (ViT). The company stated that this approach not only creates a new level of technology in terms of self-directed methods but also produces remarkable results unique to this combination of artificial intelligence methods.
Facebook also stated that its model can recognize and segment objects in images or videos without segmentation. -A targeted goal, all of which will effectively automate this process, That’s why the company claims that their features are the best.
The company also stated that object segmentation is one of the most challenging tasks in image processing because AI really needs to understand the content in the image and simplify the task from replacing the background in video chats to training robots, pictures. However, all of this is done under supervised learning and requires a lot of annotated examples.
However, Facebook’s new DINO technology can display high-precision target positioning only through supervised learning and correct architecture, which makes it easier and problem-free.
This new option will help Facebook to display and tag YouTube-like products in its video on its app, as well as other AR-related features and visual tools that may lead to more advanced and comprehensive Facebook features, which will definitely help The company integrates more data and data personalization.
Facebook is still working on this feature, and once launched, we can’t wait to know if it has similar features to YouTube. However, we know that YouTube and Facebook have always provided their best service, so we are confident that they will also provide the best service this time.