Security cameras are okay. Either the quality is too low, so they can record x hours of footage, or the quality is high, and they can record footage for y<x hours. Either way, it is inefficient.

What it does

Our program and uses facial detection as well as facial recognition to capture HD stills of peoples faces. It then compresses the video. On average, it leads to a ~85-90% decrease in file size.

How we built it

We used OpenCV, python, PIL, and Python Flask for the backend. For the UI, it was created with simple javascript calls to the Flask API, also using HTML/CSS.

Challenges we ran into

Image detection was easy, image recognition was difficult. The same person kept appearing as different people, so we had to devise a solution to train the model with limited training data. Multiple faces in the same image also produced problems for recognition.

Share this project: