FacialFeatures
-
run "make" to compile the codes
-
there are two modes for FaceFeatureDetect:
Train new data./FaceFeatureDetect -t <positive_urls> <negative_urls> <model_save_directory>
Predict a well aligned and cropped image./FaceFeatureDetect -p <hog_path> <model_directory>
Sampled data provided for training and testing:
./FaceFeatureDetect -t data/exp1_pos.hog data/exp1_neg.hog data/exp1
./FaceFeatureDetect -p data/pos.hog data/exp1 (outputs 1)
./FaceFeatureDetect -p data/neg.hog data/exp1 (outputs 0)
NOTE:
1. The train mode will not output anything.
2. The predict model will output a single number for the predicted label