Understand-MaskRCNN is a extremely simple FasterRCNN, MaskRCNN repo for explaining how it works.
-
The main purpose of this project is creating a easy to understand code, because FasterRCNN/MaskRCNN is still the strongest basline in two-stage object detection algorithm so far. I try to avoid using extra packages, so that users don't need to install and take time to learn other works.
-
Furthermore, I try to unlock every black box in the algorithm. It allows whether you are a beginner, half-understanding, or only unclear one or two points, you can fully understand it through this project.
-
Note: If you want a powerful and efficient MaskRCNN tool. Please refer to Detectron2, maskrcnn-benchmark, mmdetection (Pytorch),and Tensorflow Object Detection API , matterport (Tensorflow)
pip install numpy
pip install opencv-python
pip install torch==1.4.0 torchvision==0.5.0
Open MaskRCNN/main.py
data_path = os.path.join(os.path.dirname(os.getcwd()), 'dataset')
train_net(data_path, max_epoch = 300)
Open MaskRCNN/main.py
data_path = os.path.join(os.path.dirname(os.getcwd()), 'dataset')
inference(data_path, score_thresh = 0.8)
- Region Proposal Network
- Faster RCNN
- Mask RCNN
- The article of Understand-MaskRCNN
- Flexible anchors mask rcnn