All Tools
M
OtherFreeOpen Source
MMSELFSUP
OpenMMLab self-supervised learning toolbox and benchmark
Apache-2.0
ABOUT
Self-supervised vision methods live in incompatible paper codebases. MMSelfSup unifies contrastive, masked-image, and clustering recipes under OpenMMLab configs so teams pretrain and transfer backbones without rebuilding each method's data pipeline.
INSTALL
pip install mmselfsupINTEGRATION GUIDE
1. Pretrain vision backbones with MoCo, SimCLR, or MAE-style recipes
2. Transfer self-supervised checkpoints into MMDetection or classification
3. Benchmark SSL methods on ImageNet-style datasets with one toolbox
4. Fine-tune unlabeled-domain backbones before downstream detection
TAGS
pythoncomputer-visionself-supervisedpretrainingopenmmlabpytorchcontrastive