-
Supported for Static RGB Images
-
Pretrained Model :
https://github.com/nzx9/Hand-Sign-Recognizer/blob/main/models/pretrained_models/static_asl_rgb.pth -
Pretrained Model Used American Sign Language (ASL)
-
Pretrained Model Trained Using : https://www.kaggle.com/grassknoted/asl-alphabet
- Install python 3.6+ version
- Install pip3
- Open the
terminal
and clone the repository by running,
git clone https://github.com/nzx9/Hand-Sign-Recognizer.git
- Navigate to the project folder,
cd Hand-Sign-Recognizer
- Create virutal environment and activate the env
python3 -m env ./env
source ./env/bin/activate
- Install the requiements by running,
pip3 install requirements.txt
- Install python 3.6+ version
- Install pip3
- Open the CLI and clone the repository by running,
git clone https://github.com/nzx9/Hand-Sign-Recognizer.git
Argument | Short | Help | Default |
---|---|---|---|
--train_dataset | -trd | Path to training dataset | ./data/asl_alphabet/asl_alphabet_train |
--epochs | -e | Number of epochs | 10 |
--learning_rate | -lr | Learning rate | 1e-4 |
--batch_size | -b | Batch size of dataloader | 64 |
--num_workers | -n | Number of workers in dataloader | 2 |
--save | -s | Name of the .pth file to save | 'asl-interpreter-rgb.pth' |
--output | -o | Show Output | False |
python3 trainer.py -trd [? path_to_train_dataset] -e [? epochs] -lr [? learning_rate] -b [? batch_size] -n [? num_workers] -s [? save_to] -o [? output]
python trainer.py -trd [? path_to_train_dataset] -e [? epochs] -lr [? learning_rate] -b [? batch_size] -n [? num_workers] -s [? save_to] -o [? output]
Argument | Short | Description | Default |
---|---|---|---|
--test_dataset | -tsd | Path to testing dataset | ./data/results/asl_alphabet_test |
--batch_size | -b | Batch size of dataloader | 32 |
--num_workers | -n | Number of workers in dataloader | 2 |
--pth_file | -p | File path to .pth file | ./data/results/static_asl_rgb.pth |
--save | -s | Name of the output files need to save | test_outputs |
--confusion_matrix | -c | Show Confusion Matrix (Not Implemented Yet) | - |
--output | -o | Show output | False |
python3 tester.py -tsd [? path_to_test_dataset] -b [? batch_size] -n [? num_workers] -p [? pth_file] -s [? save_to] -c [? confusion_matrix] -o [? output]
python tester.py -tsd [? path_to_test_dataset] -b [? batch_size] -n [? num_workers] -p [? pth_file] -s [? save_to] -c [? confusion_matrix] -o [? output]
-
Training Accuracy: 99.044% (86168/ 87000)
-
Testing Accuracy: 100% (26/ 26)