Results
One of the first steps in our filtering algorithm is detecting a bounding box around the desired features. Here you can read more about how that is accomplished.
Using the Viola-Jones Algorithm we were able to apply basic filters to images by first detecting faces then appropriately scaling the filter objects and applying them to the image. We were able to place objects like hats on the top of people's heads as well as placing objects in the middle of faces like mustaches. In this gallery you can see some of our results.
Using Machine Learning techniques we attempted to train the Viola-Jones Algorithm to more accurately detect features. In this gallery you can see some of our results.
By combining our filter implementation for still images with the KLT algorithm we were able to apply filters to videos. You can see the results of this filtering in this gallery.