Disclaimer

Please note that this readme is out of date, so it may not include full instructions. Author may update it in the future.

Mask Detection and Dancers Recognition

This project aims at detecting mask location and recognize images for typical dances from the city of Cusco - Peru.

Basic system requirements:

• OpenCV 3.0.0 or greater
• Numpy 1.11.2
• Scikit-image 0.12.3
• Scikit-learn 0.18

Getting the labels

getLabels.py: Divides the original images into small pieces and labels them with the percentage of the containing face. The dataset file must contain ~/Path-to-original-dataset/rgb and ~/Path-to-original-dataset/bin subdirectories. Put original images in ~/Path-to-original-dataset/rgb and the binary images (ground truth) in ~/Path-to-original-dataset/bin. The pieces will be saved in a /patches directory at same hierarchy of ~/Path-to-original-dataset, a subdirectory for each type of w, h, pyr_hight and shift is created.

To execute the script, run:

python getLabels.py -d ~/Path-to-original-dataset

Training the models

train.py: Uses the results of getLabels.py to train a model (SVM, Random Forest, AdaBoost) and save the learned parameters in /models directory at same hierarchy of the ~/patches directory. A subdirectory for each w, h, pyr_hight, shift and descriptor parameters is created.

Run it with:

python train.py -d ~/patches/directory-with-small-labeled-images

##Running a test

detect.py: Reads the trained models, and for each image in the directory plots an image with the face locations.

Run it with:

python detect.py -m ~/models/directory-with-learned-models -t ~/path-to-test-directory