从2018年年中开始,香港中文大学多媒体实验室(MMLab)就启动了OpenMMLab(https://github.com/open-mmlab)计划。这项计划的初衷是为计算机视觉的一些重要方向建立统一而开放的代码库,并不断把新的算法沉淀其中。我们相信,这一项工作可以推动可复现算法生态的建立,也是对计算机视觉社区的重要贡献。
主要项目包括:
图像分割-mmsegmentation,OpenMMLab Semantic Segmentation Toolbox and Benchmark.
视觉基础-mmcv, OpenMMLab Computer Vision Foundation.
图像视频编辑-mmediting,OpenMMLab Image and Video Editing Toolbox.
动作理解2-mmaction2,OpenMMLab's Next Generation Action Understanding Toolbox and Benchmark
3维目标检测-mmdetection3d, OpenMMLab's next-generation platform for general 3D object detection.
自监督学习-OpenSelfSup,Self-Supervised Learning Toolbox and Benchmark
服装时尚分析-mmfashion,Open-source toolbox for visual fashion analysis based on PyTorch
动作理解-mmaction,An open-source toolbox for action understanding based on PyTorch
光流提取-denseflow,Extracting optical flow and frames
图像分类-mmclassification,OpenMMLab Image Classification Toolbox and Benchmark
LiDar目标检测-OpenPCDet,OpenPCDet Toolbox for LiDAR-based 3D Object Detection.
姿势估计-mmpose,OpenMMLab Pose Estimation Toolbox and Benchmark.
非监督自适应对象识别-OpenUnReID,Open-source toolbox for unsupervised or domain adaptive object re-ID.
人体姿势估计-mmskeleton, A OpenMMLAB toolbox for human pose estimation, skeleton-based action