Jake Ireland

I should also mention [this is the data](https://github.com/opencv/opencv/blob/4.x/data/haarcascades/haarcascade_frontalface_default.xml) we want to read from.

I think I need to do some reading to answer some questions about the structure of the XML data[[1]](https://www.uni-augsburg.de/de/fakultaet/fai/informatik/prof/mmc/research/publications/#opus-year-2003), [[2]](https://opus.bibliothek.uni-augsburg.de/opus4/frontdoor/index/index/docId/67686), [[3]](https://opus.bibliothek.uni-augsburg.de/opus4/frontdoor/index/index/docId/67695)

The difference in faceness scores between positive and negative images seems to be statistically significant after some changes, just using 1000 training images and 1000 testing images: [scores.pdf](https://github.com/jakewilliami/FaceDetection.jl/files/7501345/scores.pdf)

Could learn from [this](https://au.mathworks.com/help/vision/ref/vision.cascadeobjectdetector-system-object.html) or [this](https://github.com/aldebaro/ufpa-face-detection).

More work on this in 52ce320b

Possibly use methods from [`YOLO.jl`](https://github.com/IanButterworth/YOLO.jl) for this?

Maybe [this](https://github.com/salvacarrion/viola-jones/blob/44783aab741dbca2630c88bd59b14516e3d6c6b4/violajones.py#L117-L154) is a thing

Adding LICENSE (f356de1f) and VERSION (38b6e987) is a step towards this

A step towards this was made in fa4352ce

A step towards this (brought to my attention by @dmipeck) would be to not use absolute paths (rather, dynamic paths relating to where the file is being run from; e.g.,...