Overview
Based on Tencent's industry-leading AI technology, VOD’s face recognition feature helps you quickly recognize faces in a video and provides information including the frames a face appears in and the coordinates of the face frame. You can use VOD’s public person libraries or customize and manage your own person libraries. The public person libraries contain celebrities in various fields, such as movies, music, sports, and academics.
Use cases
Face recognition in media content has been widely used for video creation, media search, and generation of personalized recommendations.
|
Video creation | The face recognition capability increases the efficiency of content production by allowing users to quickly locate specific people from a large number of videos and obtain information including when a person starts to appear, where the person appears on an image, and how long the appearance lasts. |
Event replays and highlights | You can add timestamps for when a certain player appears during a sports and esports event, so users can quickly locate the relevant time points. You can also find the time points and capture the video frames in which a certain celebrity appears in a video, making it easier to create a highlight reel or generate an animated image as the thumbnail to attract more viewers. |
Video websites | Users can search for relevant videos by the name of a person. Video websites can also recommend videos based on the list of persons followed by a user. |
How to Use
Face recognition is enabled by the face recognition capability of Video Content Recognition. To use the feature, follow the steps below: 1. Prepare an audio/video content recognition template.
You can use a preset audio/video content recognition template to recognize all people in the public person libraries.
You can also create your own template and specify the categories of faces in public or custom person libraries to recognize by setting the FaceConfigure parameter. For example, the following code sample indicates to recognize only sports celebrities in the public person libraries: {
"FaceConfigure": {
"Switch": "ON",
"DefaultLibraryLabelSet": ["sport"]
}
}
Get the template ID from the response.
* You can manage custom person libraries using the AI-based Sample Management APIs (CreatePersonSample/DeletePersonSample/DescribePersonSamples/ModifyPersonSample).