Neural Networks are great at discerning patterns from structured and unstructured data alike. One of the most interesting areas of application of neural nets is in the field of Computer Vision. Convolutional neural nets can allow computer to see, identify and separate objects within their field of vision and learn to react accordingly.
In this challenge, you will be tasked to create a module that can extract facial data out of a video feed and recognize each person individually in the video. Your only constraints are that you are to use Python and you cannot use a already baked-in solution! You will have to use your best Google-fu to find the best resources out there to speed up your development process. You are allowed to use already trained neural networks, or if you are up for the challenge, train your network yourself! We highly recommend the first path though ;)
Thus here are the milestones. The milestones are mostly there to give you an idea of your progression, but they may differ from your tutorials'. In this case don't worry and keep up your good work 🚀
This means!
- You have trained/found an already trained network
- You have tested the network against a test set and it has been successful (90%+ success is a good benchmark)
Aka Milestone 1A for videos :)
You must now build an application that can track people in a video.
Now you will combine Part 1 & Part 2! You have to build an application that can track people once they have been identified using their face :O
Put your skills to the test! Free your creativity! From here on out, you decide what you do! Here are some ideas:
- Sentiment analysis
- Natural Language Understanding for sound
- Integrate support for camera + web interface
- Integrate hardware (e.g. Raspi)
- Statistical analysis (time in frame, number of people, etc)