I have to detect faces in videos using machine learning methods like SVM, I have to build it from scratch not using any inbuilt face detectors.My time limit is like 15hours from now.Can I get any valuable suggestions on how can I begin and proceed? Thank you
If you only have 15 hours I would built it on a well known easy to understand algorithm like Viola and Jones, which uses weak binary classifiers and combines them to detect faces. It is relativily fast algorithm that runs in realtime.
You can get more information on the algorithm in the following stack exchange Post