Video analysis

Prev Next

The latest service changes have not yet been reflected in this content. We will update the content as soon as possible. Please refer to the Korean version for information on the latest updates.

Available in VPC

Video analysis is the process of analyzing video assets using the multimodal engine (Sinossi) and extracting metadata.
To generate an index by analyzing a video:

  1. Access the Media Intelligence workspace.
  2. Add the video file to be analyzed to your project.
  3. Click the [Start analysis] button of the video to be analyzed.
    • Alternatively, click the check box for the target video, then click the [Start analysis] button at the top of the screen.
  4. Configure the following settings in [Set analysis options] of the start analysis window:
    Item Description
    Analysis engine
    Scene length
    Analysis language
    Voice effect detection
  5. Configure the following settings in [Set analysis person] of the start analysis window:
    Item Description
    Person detection
    Person tag
  6. After all settings are completed, click the [Start analysis] button.
    • Once the analysis is completed, the index ID appears in the analysis results.
    • To cancel the analysis, click the [Cancel analysis] button.
  7. To perform additional analysis, click the [Add analysis] button.
    • You can analyze a video asset multiple times.
Note
  • When setting the scene length, select Long for genres such as documentaries or dramas where dialogue is long and context is important, and Short for genres such as reality shows where scene changes are frequent.
  • If confidence level filtering is set to high, you can obtain highly accurate results, but the chance of missing detections increases.
  • If confidence level filtering is set to low, you can obtain more detection results, but the chance of detecting unnecessary objects increases.