Project 5 / Face Detection with a Sliding Window

Multi-scaling with Stepsize of 6

In this project, we will use machine learning to help us detect faces in images. To do so, we first need features to give to our model for it to train, then we would train our model and finally use this on a test dataset of images to find faces. To get the features, we used Histogram of Oriented Gradients (HOG). To train our classifier, we used the features to train our model to detect faces. For this task, I used Support Vector Machine with a lambda of 0.0001. Finally, we are now able to use our model as a sliding window on test images at multiple scales.

Face template HoG visualization with step size 6
Precision Recall curve for step size 6
Face template HoG visualization with step size 4
Precision Recall curve for step size 4








































Results with step size 6