What should be suggested way to do object classification in a streamed video using pre-trained model with custom label?

Question

Hello Everyone,

I'm trying to create new model or with the help of pre-trained model for classification which can support video streaming data.

I've tried custom label in AWS Rekognition for classification but I could evaluate that model only with image data (not in a streaming mode). Is there a way to evaluate a model for object detection or classification in a streamed video data?

I've used AWS kinesis video stream to stream my laptop cam to AWS kinesis video stream. When I try to integrate Kinesis video stream with AWS Rekognition, I understood that it supports only face detection and specific object detection like pet, person or package labels with the pre-trained model. Is my understanding correct? or is there a way to use pre-trained model for a custom labels in a video streaming data.

Can anyone suggest me the better solution for this requirement?

Any help would be appreciated.

tranja · Answer

Hello,

It seems from your question that that Rekognition Custom Labels is what you're looking for, but the only feature that is lacking is the ability to stream video. One option is to extract images from Kinesis Video Streams ([here](https://docs.aws.amazon.com/kinesisvideostreams/latest/dg/images.html)) and pass those images to Rekognition Custom Labels.

If you want to use off-the-shelf Rekognition models and directly stream video you are limited to either [face detection](https://docs.aws.amazon.com/rekognition/latest/dg/collections-streaming.html) or [person, pet, package detection](https://docs.aws.amazon.com/rekognition/latest/dg/streaming-video-detect-labels.html). Please let me know if you have anymore questions.

What should be suggested way to do object classification in a streamed video using pre-trained model with custom label?

Relevant content