A face mask detector built with Deep Learning and OpenCV
Objectives of this project are to build a detector to detect whether a person is wearing a mask and further evaluate the performance of this detector.
This detector is built using PyTorch, transfer learning is implemented with model from FaceBoxes: A CPU Real-time Face Detector with High Accuracy and evaluation of the model performance is performed using mAP (mean Average Precision)
Install:
pip install pytorch
pip install opencv-python
Python 3.8.5, PyTorch v1.5.0 and OpenCV v4.0.1 are used.
To do transfer learning, pre-trained model can be downloaded here and place this in pretrained_model/ folder. This pretrained-model is shared by authors of FaceBoxes in their repo.
Annotated datasets can be downloaded from AIZOOTech, data are downloadable in both GoogleDrive and BaiduDisk.
Navigate to face_mask_detector/ then run:
python train.py
To evaluate the trained model, run:
python evaluate_detector.py