News: SBLDet: Scale Balanced Learning for object Detection.This novel method is based on mmdetection to reimplement Scale Balanced Learning Moudle and use the model Faster R-CNN HBB to get the results. The master branch works with PyTorch 1.1 or higher. mmdetection is an open source object detection toolbox based on PyTorch. It is a part of the open-mmlab project developed by Multimedia Laboratory, CUHK.
Supported methods and backbones are shown in the below table. Results and models are available in the Model zoo.
| ResNet | ResNeXt | SENet | VGG | HRNet | |
|---|---|---|---|---|---|
| RPN | ✓ | ✓ | ☐ | ✗ | ✓ |
| Fast R-CNN | ✓ | ✓ | ☐ | ✗ | ✓ |
| Faster R-CNN | ✓ | ✓ | ☐ | ✗ | ✓ |
| Mask R-CNN | ✓ | ✓ | ☐ | ✗ | ✓ |
| Cascade R-CNN | ✓ | ✓ | ☐ | ✗ | ✓ |
| Cascade Mask R-CNN | ✓ | ✓ | ☐ | ✗ | ✓ |
| SSD | ✗ | ✗ | ✗ | ✓ | ✗ |
| RetinaNet | ✓ | ✓ | ☐ | ✗ | ✓ |
| GHM | ✓ | ✓ | ☐ | ✗ | ✓ |
| Mask Scoring R-CNN | ✓ | ✓ | ☐ | ✗ | ✓ |
| FCOS | ✓ | ✓ | ☐ | ✗ | ✓ |
| Double-Head R-CNN | ✓ | ✓ | ☐ | ✗ | ✓ |
| Grid R-CNN (Plus) | ✓ | ✓ | ☐ | ✗ | ✓ |
| Hybrid Task Cascade | ✓ | ✓ | ☐ | ✗ | ✓ |
| Libra R-CNN | ✓ | ✓ | ☐ | ✗ | ✓ |
| Guided Anchoring | ✓ | ✓ | ☐ | ✗ | ✓ |
Other features
- DCNv2
- Group Normalization
- Weight Standardization
- OHEM
- Soft-NMS
- Generalized Attention
- GCNet
- Mixed Precision (FP16) Training
- Please refer to INSTALL.md for installation and dataset preparation.
- Before install, you should make sure the configuration is correct
vim ~/.bashrc
PATH="/mnt/lustre/share/gcc/gcc-5.3.0/bin:$PATH"
export CC="/mnt/lustre/share/gcc/gcc-5.3.0/bin/gcc"
export CXX="/mnt/lustre/share/gcc/gcc-5.3.0/bin/g++"
vim ~/.condarc
channels:
- https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/
- https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
- https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/
- https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/msys2/
- https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
show_channel_urls: true- You can install directly from the script below
export INSTALL_DIR=$PWD
mkdir checkpoints
mkdir data
mkdir work_dirs
conda create -n mmlab python=3.7 -y
source activate mmlab
conda install pytorch torchvision==0.2.2 cuda90 cudatoolkit=9.0 -y
conda install cython -y
git clone [email protected]:yanhongchang/mmdetection.git
cd mmdetection
git checkout horizontal
python setup.py build develop
ln -s ../data data
ln -s ../checkpoints checkpoints
ln -s ../work_dirs work_dirs
# rm -rf /mnt/lustre/yanhongchang/.conda/envs/open-mmlab/lib/python3.7/site-packages/torchvision-0.4.1-py3.7-linux-x86_64.egg/
unset INSTALL_DIR- Please see GETTING_STARTED.md for the basic usage of MMDetection.
- Data processing
- Preare data and checkpoints.
- run scripts
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent