Filters using computer vision

This is a project for ACM's workshop

Notes

• Dark png's tend to mess up the masking so they are avoided.
• Code is based off of tutorials found. e.g. https://medium.com/@mohd-uzair/making-face-filters-with-opencv-e3c928865239

Current features

• Cycle filters using 0-9 keys
• Debug prints

TODO

• EMA smoothing for less twitchy filter application
• More scalable code (have to add case for each new png)
• Fix hat filters. They are generally out of the detected face coords.

Bugs

• Program is very prone to curvature. A camera tilt can make the filter look bad. I'm not aware how I can fix lens distortion.
• Fix out of bound errors
• If the user is too far away, scaling factor will be too much and mess up filter positioning