This repository is fork of https://github.com/mrnugget/opencv-haar-classifier-training . The details are on this blog. Please visit this blog or the original repository for details. The official OpenCV documentation is pretty good too https://docs.opencv.org/3.4.3/d7/d8b/tutorial_py_face_detection.html
Put negative images into the folder ./negative_images/ and positive images
into ./positive_images/. And run
make
To use the existing images, you need to install git-lfs and run
git lfs pull
make
to pull the default images and start the process.
It will create a LBP classifier and save it to ./classifier_lpb/cascade.xml.
To create a HAAR classifier (It will at least 100 times more time to train), run
make haar and it will save the classifier to ./classifier_haar/cascade.xml.
If you have your recording in TIFF file, you can use test_cascade.py file
e.g.,
python ./test_cascade.py --tiff trial_008.tif --cascade ./trained_classifiers/mouse_eye.xml
And it will locate the trained pattern.