amazon ec2 - How can I use EC2 instance for real time inference?

Question

I want to be able to run deep learning model on EC2 server with webcam as input real time.The entire process is described in following steps:

1. Get webcam feed
2. Transfer frames to EC2 instance
3. Run inference of a deep learning model(An alpha matting model) and get the output
4. Transfer the output back to client(local machine with webcam) and display

This process should ideally be almost real-time with minimal lag. The model runs around 20fps.

Answer 1

等待大神答复