Computer Vision

Question

I wanted to know that how to detect specific activity(like eating , drinking etc) in a video stream using python or c++.Are there any libraries for such things?

Answer 1

You'll mostly likely require deep learning. Get a lot of pictures of people (not) eating or drinking, train a neural network to understand when a picture describes a particular activity and when it doesn't. When the model is sufficiently trained, test it on your video by splitting the video to frames and feed the frames to the model to obtain a confidence level / output.

If you want to go deep, I can explain the steps you need to take to preprocess the images, represent the image data as a matrix to be used as training data, etc... Or, you can just try out readily available samples:

https://cloud.google.com/vision/
https://cloud.google.com/automl/

Answer 2

melip