GUIDE
- To train, run the function TrainHog()
eg. TrainHOG('parameters.txt', 'data', 'weights.txt');
- To predict one image run the function PredictHOG()
eg. PredictHOG('weghits.txt', 'data/test/neg_6370', 'parameters.txt');
3)Extra: to predict multiple images in a folder, run Predict
eg. Predict('weights.txt', 'data', 'parameters.txt');
IDEAL SET UP
1)positive images: \data\pos\
2)negative images: \data\neg\
3)testing images: \data\test\
PARAMETERS FILE
1 Image Resize Height 2 Image Resize Length 3 Cell Size 1 4 Cell Size 2 5 Number of Bins 6 Block Size 1 7 Block Size 2 8 Overlap 1 9 Overlap 2 10 Feature Vector Size (not used by default) 11 Learning Rate
must be comma separated values
e.g.: 128, 64, 9, 9, 16, 5, 5, 3, 3, 10000, 0.04
result.txt format:
Negative score, Positive score