Project 1
Algorithmically determine if a picture contains people
Jones, Jonathan
This program performs facial feature detection on any given directory of images. You can optionally copy the detected images into a new location if desired. If you’ ve ever heard of Google’ s Vision API, this program can be thought of as a mini version of that.
The original idea to implement a program like this came about when I was filtering through old hard drives searching for some family photos. Using this program for the 100+ thousand images I came across was the only reasonable approach to finding the pictures I was looking for.
After running the program with the directory of example images, the final results are shown to be acceptable. Shown below is a listing of the example images followed by a listing of images that the program determined to contain a person.
We can see that the program is not 100% accurate, and it’ s unreasonable to expect that kind of accuracy. However, it detected 4 images correctly and 2 incorrectly (0% false negatives; 6% false positives).
The source code for this program is also hosted on GitHub at the following link:
https://github.com/jjones646/humfilt
The example images were taken from the source code for OpenCV, which is licensed under a BSD license.
The classifier tree used for feature detection was also taken from the OpenCV source code.
Another notable area of credit is the official OpenCV tutorial about facial detection.
