Video Frame Extraction Procedure

Question

Hello.

I am attempting to extract a single frame from a video where there is an object in focus. However, sometimes there can be person in the view of the video. I only want to extract and output a single frame from the video without the human obstruction.

My initial thought process on how to do this is with Amazon Lambda. However, after doing more research, it seems that I also have the option to use AWS Batch and Amazon ECS. I am in a dilemma of which one to use.

I simply want to extract a single frame from videos that are stored in S3 as mp4 format. The dataset is also fairly large for context. I also hope to utilize Amazon Rekognition to help detect any human that may be in the video and not extract the single frame when the person is in frame.

Is this the right way to approach this procedure? Also, which tool would be the best to use (Lambda, Batch, or ECS)?

Didier Durand · Accepted Answer

Hi,

AWS Batch is more applicable to one-off process like the one you describe. ECS / Fargate is more aimed at permanent back-ends with which multiple requesters will interact on a permanent basis.

On my side,  I used AWS Batch for your kind of job.

BTW, as you will notice, AWS Batch indeed relies on generic pools of ECS / Fargate servers (managed by AWS Batch "under the hood"). If you choose the Fargate option in your Batch setup, you will rapidly have your batch up & running.

Best,

Didier

Jay Busch · Answer

Thanks for your question shimekan@. We covered this question in our recent episode of re:Post Live. Check it out here: https://repost.aws/articles/ARopx6otB8SEqK_ivqXC7cEQ/join-us-live-on-tuesday-september-2nd-for-aws-re-post-live-gen-ai-for-advertising-decoding-the-art-of-the-possible

Video Frame Extraction Procedure

Add your answer

Relevant content