I need to build a multi-camera people tracking system and I have no idea how to start. I read ML for Dummies and I've watched a lot of youtube classes/conferences and read a lot of articles about ML/DL, so I have all this theoretical information about what is a NN, loss function, weights, vectors, convolution, etc., but when I need to start building something, I get stuck. Even more, I don't think I can create my own models because I only have six months to finish this and I'm not sure if I'll be able to do it.

I've read some papers explaining architectures for an improved people-tracking system (e.g. https://www.intechopen.com/online-first/multi-person-tracking-based-on-faster-r-cnn-and-deep-appearance-features#B8), and it says it used ResNet-30 and stuff like that. My question is, how could I recreate the architectures in papers like that? Where can I find those pre-trained models? Or is there a place where I can get the data?

I want to start with at least a people-tracking system, without worrying about the multi-camera part for now, and I thought of almost the same approach as the people in the paper posted, meaning I want to recognize people based on parts of their body/the whole body to identify them, and track them based on their unique features (clothing color, hair, skin tone, etc), maybe skipping the part of facial recognition since that's too advanced I think.

Any idea on where to start? Sorry if the question is too broad or too complex. Comments about first steps and sub-dividing the problem are also welcome. PS: The main ultimate is to track how much time people are in a certain area filmed by many cameras.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.