OpenMMLab builds the most influential open-source computer vision algorithm system in the deep learning era. It aims to
- provide high-quality libraries to reduce the difficulties in algorithm reimplementation
- create efficient deployment toolchains targeting a variety of backends and devices
- build a solid foundation for computer vision research and development
- bridge the gap between academic research and industrial applications with full-stack toolchains
Based on PyTorch, OpenMMLab develops MMEngine to provide universal training and evaluation engine, and MMCV to provide neural network operators and data transforms, which serves as a foundation of the whole project. Since the initial release in October 2018, OpenMMLab has released 30+ vision libraries, has implemented 300+ algorithms, and contains 2000+ pre-trained models.
