0

I am new to the Video Intelligence API, but in my project, I am presented with the challenge of taking a video feed and running it through the video intelligence API then screenshotting it each scene change and reading text from those screenshots through the vision API. Is there a way I can do this with some sort of live stream? Thank you!

John Johnson
  • 1
  • 1

2 Answers2

1

As per documentation, live-streaming video processing is not available. Documentation only covers support for .MOV, .MPEG4, .MP4, and .AVI video files, stored locally or in the GCS.

Let me provide some information that it may help you.

Video Intelligence lists a set of features available for video analysis. One of them is Label Detection to identify entities, similar to what Vision API does, just keep in mind that Video Intelligence has not all the features that Vision has. In the other hand, I don't think you can screenshoting to read text (there is not a feature for that), instead you can only detect shot changes. An example for Label Detection:

curl -s -H 'Content-Type: application/json' \
    -H "Authorization: Bearer "$(gcloud auth application-default print-access-token) \
    'https://videointelligence.googleapis.com/v1/videos:annotate' \
    --data "{ \
   'inputUri':'gs://bucket-video/myvideo.mp4', \
   'features':['LABEL_DETECTION'] \
}"

Don't forget to set GOOGLE_APPLICATION_CREDENTIALS=/JSON/CREDENTIALS/PATH

rsantiago
  • 2,054
  • 8
  • 17
1

As of 2020 you can do that using the Beta features present in Google Video Intelligence APIs: https://cloud.google.com/video-intelligence/docs/streaming/live-streaming-overview

Aniruddha Pandey
  • 428
  • 6
  • 9