People can easily recognize the walking style of their friends. Let's see if we can teach a computer to identify walking people on videos.
What it does
It identifies people on videos based on their walking style.
How we built it
We build a neural network in torch and trained in on Amazon Web Service because we needed a lot of computing power. Additionally, we've made a website client in Ruby on Rails and a back-end server in Node.js which runs a trained neural network. When this neural network receives a video it returns the probabilities for every person it knows that the person is walking in the uploaded video.
Challenges we ran into
- Training a neural network is reeeally computing power demanding. We had some tough times trying to connect to AWS servers.
- Connecting all the servers together.
- Choosing a proper model for the neural network.
- We had to record a massive amount of videos of walking people.
Accomplishments that we are proud of
Well, we made an awesome neural network based project! Yay!
What we learned
We learnt a lot about the technologies we used.
What's next for tundra-force team
We're going to...