Our goal was to design a classifier that, given an image, is able to understand and recognize the American sign language (ASL) and output the right character represented by the sign.
conda create --name asl python==3.7.11
conda activate asl
pip install -r requirements.txt
$ cd dataset
$ python dataset_download.py
python main.py --data-aug DATA-AUG --target-size TARGET-SIZE --epochs EPOCHS --batch-size BATCH-SIZE --lr LR --name-model NAME-MODEL --fine-tune FINE-TUNE --only-test ONLY-TEST
where:
DATA-AUG: data aumentation, default is TrueTARGET-SIZE: target size of the image, default is (100, 100)EPOCHS: number of epochs, default is 50BATCH-SIZE: batch size, default is 64LR: learning rate, default is 0.0001NAME-MODEL: name of model, defaul is VGG19FINE-TUNE: fine tuning, default is FalseONLY-TEST: only test, default is False
for example, for training
python main.py
for testing:
python main.py --only-test True
ASL.mp4
