In NodeJS, I need to run some Vision API call on a live video stream.
What pre-processing to do to select the best frame before calling the API ?
- Compare frame N with N-1 ? How ?
- Detect motion blur ?
- What frame rate should I use ? 1 fps ?
I didn't find best practice to extract frames from video for SaaS Computer Vision processing. It seems to be a common use case ? Video are too large to upload.