conda create --name maskrcnn
conda activate maskrcnn
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/menpo/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/bioconda/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/msys2/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/
conda install pytorch torchvision cudatoolkit=10.0
tips: if pycocotools can not successfully be installed, download *.whl from https://pypi.tuna.tsinghua.edu.cn/simple/pycocotools-windows/
pip install /path/to/pycocotools_windows_xxxx.whl
convert your own dateset from labelme format to PennFudanPed format you can reference to this video: https://www.bilibili.com/video/BV1R7411F7QP
- Put your *.jpg files and json files together in one file after annotating, then run
python new_json_to_dataset.py path/to/original/dataset
- Extract png and gth
put copy.py together with *.jpg and json, then run
python copy.py
- Copy your pngs and masks into PennFudanPed2/PNGImages ans PennFudanPed2/PedMasks manually
- before training, edit the parameters on line 141, 184, 185 in train.py to your needs.
- names = {'0': 'background', '1': 'train'}
- num_classes = 2
- train_num = 1200
- Start training
python train.py
before predicting, editing the same params as training.
python predict.py
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent