All Products
Search
Document Center

ApsaraVideo VOD:Video fingerprinting

Last Updated:May 09, 2024

ApsaraVideo VOD provides the video fingerprinting feature based on video recognition technologies. This allows you to uniquely identify video files. Video fingerprints remain unchanged regardless of whether you convert file formats, edit and merge video files, or perform transformations such as compression and rotation. You can use the video fingerprinting feature in scenarios such as duplicate video identification and copyright protection. This topic describes how to automatically trigger a video fingerprinting job or call an API operation to create a video fingerprinting job.

Highlights

In most cases, a video fingerprint is a unique binary string that identifies a video. You can use the video fingerprinting feature to extract and compare fingerprint characteristics of images and audio in videos. This allows you to find duplicate videos, trace the source of video clips, and identify video plagiarism.

  • Uniqueness: Each video has a unique fingerprint.

  • Stability: A video fingerprint remains unchanged regardless of whether you convert file formats, edit and merge files, or perform transformations such as compression and rotation.

Scenarios

  • Duplicate video identification

    You can create an independent video fingerprint library to compare the fingerprints of sample videos against the library to find and remove duplicate videos or clips in real time. This prevents excessive identical or similar videos from being pushed to users and affecting user experience during content delivery and personalized recommendations.

  • Plagiarism identification

    The video fingerprinting feature allows you to accurately identify whether a video uses original content or plagiarizes other creators. This provides technical support for copyright protection and helps build a healthy video ecosystem in which the rights of video producers and short video platforms are protected.

  • Quick content moderation

    You can create a video blacklist to compare the fingerprints of new videos against the blacklist to determine whether the new videos contain prohibited content. Compared with traditional video moderation methods, this method can process a large number of videos with higher efficiency at lower costs.

  • Ad revenue sharing

    You can use the video fingerprinting feature to monitor and identify specific ads from an ad library for revenue sharing and easily control the time and frequency to deliver ads. This feature is beneficial to a dynamic ad revenue sharing ecosystem and protects the interests of both ad copyright owners and advertising platforms. You can also identify and replace ads, and auction ad slots in an efficient manner.

  • Video source tracing

    You can retrieve a new video from a large video fingerprint library to determine the propagation method of the video. This helps trace video sources, analyze the distribution method of videos, and explore relationships between media resources.

Before you begin

  • After you read the usage notes described in the Overview topic, submit a ticket to enable the video fingerprinting feature.

  • The first time you use the video fingerprinting feature, submit a ticket to apply to use the media fingerprint library free of charge. Otherwise, you cannot use the video fingerprinting feature as expected.

Billing

If you enable and use the video fingerprinting feature, you are charged based on the length of the processed videos. For more information about billing, submit a ticket to contact Alibaba Cloud.

Use the video fingerprinting feature

  • Triggered by the system

    After you enable the video fingerprinting feature, the system automatically creates a video fingerprinting job after a video is uploaded and transcoded in ApsaraVideo VOD. You can obtain the job results from the AIMediaDNAComplete callback.

  • Triggered by API operations

    • After you upload a video to ApsaraVideo VOD, you can call the SubmitAIJob operation to submit a video fingerprinting job for the video.

    • After a video fingerprinting job is submitted, the job is asynchronously processed in the backend. You can obtain the job results from the AIMediaDNAComplete callback. You can also call the ListAIJob operation to query job results.

      Note
      • To automatically create a video fingerprinting job, you must transcode the video. You need to export only one transcoded video stream, such as an MP4 stream in the low definition or original quality, to automatically create a video fingerprinting job.

      • If you call API operations to submit video fingerprinting jobs, you do not need to transcode the video. You can directly process source files that are not transcoded.

  • Query results

    • After a video fingerprinting job is complete, you can call the GetMediaDNAResult operation to query the job result.

    • Description about the comparison result of video fingerprints

      The comparison results of video fingerprints contain the following information:

      • Whether duplicate videos are found

      • Information about the duplicate videos

      • Interval of the timeline of a video during duplicate content from another video is shown.

      • Degree of duplication between a video and another video. The value is a number ranging from 0 to 1.

  • Existing videos

    To ensure the comparison performance, you can import existing videos to build a complete feature library. You can process existing videos by using the following methods.

Note

You cannot create an independent feature library in ApsaraVideo VOD. You can use only the default fingerprint library. To create a new fingerprint library, use ApsaraVideo Media Processing (MPS). For more information, see CreateFpShotDB.

  • If the total duration of your existing videos does not exceed 1 million minutes, you can call the SubmitAIJob operation to submit a request to extract and compare the video fingerprints of existing videos.

  • If the total duration of your existing videos exceeds 1 million minutes, submit a ticket and describe the quantity, duration, and storage location of the existing videos.