- The automated review service of ApsaraVideo for VOD is realized based on a large amount of labeled data and deep learning algorithms. This service precisely identifies prohibited content (pornographic, terrorism, and politically sensitive content) in videos, thumbnails, and titles from multiple dimensions, such as voice, text, and vision. This service applies to multiple scenarios, such as short video review and media review, and greatly improves video review efficiency. Before using this service, ensure that you have completed the preparations for media review.
For a video that has been uploaded to the media library in ApsaraVideo for VOD, you can submit an automated review job by using the API or SDK. For more information, see SubmitAIMediaAuditJob.
After the automated review job is completed, an event that notifies you of the job result is sent to the specified callback URL. For more information, see AIMediaAuditComplete. You can also query the job result through the API. For more information, see GetAIMediaAuditJob.
After automated review is completed, you can query the review result summary by using the API. For more information, see GetMediaAuditResult.
After automated review is completed, you can query the review result details by using the API. For more information, see GetMediaAuditResultDetail.
After automated review is completed, you can query the review result timeline by using the API. For more information, see GetMediaAuditResultTimeline.
Note: The image resources in the review result are stored in the free storage provided by ApsaraVideo for VOD for
two weeks. After two weeks, they are automatically deleted.
Identifies pornographic and sexual content in video thumbnails, titles, and content from multiple dimensions such as voice, text, and vision by using AI technologies.
Identifies terrorism and politically sensitive content from the following six dimensions: weapon, sensitive figure, bloody scene, specific costume, smoke and light scene, and special symbol. This helps reduce security risks such as terrorism and political turmoil.
Identifies text, watermarks, and QR codes in videos to intelligently detect advertisements placed by Micronet suppliers and advertisements inserted into videos.
Identifies undesirable scenes such as picture-in-picture, meaningless images, and smoking. An independent model is used for each scene. This resolves the mutual interference issue that occurs when a single model is used for training different scene categories. In this way, the precision and recall of identification are significantly improved.
- Short video or UGC video service providers can use the automated review service to quickly identify prohibited content in a large number of videos uploaded by users. This effectively reduces the manual review workload.
- VOD service providers and media companies can use the automated review service to efficiently identify terrorism and politically sensitive scenes in videos to strictly control prohibited content.