Hello welcome to my face detection,
Its pytorch implementation of face detection system using yolo techniques on the famous widerface
dataset : http://shuoyang1213.me/WIDERFACE/
-
Install PyTorch : https://pytorch.org/
-
Clone the Repo :
$ git clone https://github.com/AbdallahOmarAhmed/face-mask-detection
-
Install requirements :
$ pip install -r requirements.txt
-
Setup evaluation :
$ python3 eval/setup.py build_ext --inplace
- Download widerface dataset in the project dir :
$ gdown https://drive.google.com/uc?id=15hGDLhsx8bLgLcIRD5DhYt5iBxnjNF1M
$ gdown https://drive.google.com/uc?id=1GUCogbp16PMGa39thoMMeWxp7Rp5oM8Q
$ wget shuoyang1213.me/WIDERFACE/support/bbx_annotation/wider_face_split.zip
- Unzip the dataset files:
$ unzip WIDER_train.zip
$ unzip WIDER_val.zip
$ unzip wider_face_split.zip
- Run the train.py file :
$ python3 train.py [batch size]
backbone | epoch time (rtx-2070) | easy accurace | medium accurace | hard accurace | resolution |
---|---|---|---|---|---|
resnet18 | 151 sec | 88.4% | 83.4% | 56.7% | 448 * 448 |
-
Make predictions :
$ python3 make_prediction.py [your model path]
-
Test predictions :
$ python3 eval/evaluation.py
$ python3 test_image.py [model path] [image path]