windows mmsegmentation-master环境百分百成功

win10，pycharm（python3.6），Anaconda，显卡1650ti，CUDA11.2(低于11.2也可)，torch1.9.1,torchvision0.10.1，mmcv-full1.4.7

不同版本一直试错，配了三天，终于配置好了。浅浅记录一下。

1.pycharm,anaconda安装

转自

https://blog.csdn.net/qq_32892383/article/details/116137730?ops_request_misc=%257B%2522request%255Fid%2522%253A%2522166428995116782428634244%2522%252C%2522scm%2522%253A%252220140713.130102334..%2522%257D&request_id=166428995116782428634244&biz_id=0&utm_medium=distribute.pc_search_result.none-task-blog-2~all~top_positive~default-1-116137730-null-null.142^v50^pc_rank_34_2,201^v3^control_2&utm_term=pycharm%E5%92%8Canaconda%E5%AE%89%E8%A3%85%E6%95%99%E7%A8%8B&spm=1018.2226.3001.4187https://blog.csdn.net/qq_32892383/article/details/116137730?ops_request_misc=%257B%2522request%255Fid%2522%253A%2522166428995116782428634244%2522%252C%2522scm%2522%253A%252220140713.130102334..%2522%257D&request_id=166428995116782428634244&biz_id=0&utm_medium=distribute.pc_search_result.none-task-blog-2~all~top_positive~default-1-116137730-null-null.142%5Ev50%5Epc_rank_34_2,201%5Ev3%5Econtrol_2&utm_term=pycharm%E5%92%8Canaconda%E5%AE%89%E8%A3%85%E6%95%99%E7%A8%8B&spm=1018.2226.3001.4187

2.mmsegmentation下载

GitHub - open-mmlab/mmsegmentation: OpenMMLab Semantic Segmentation Toolbox and Benchmark.

3.CUDA安装

转自https://blog.csdn.net/m0_45447650/article/details/123704930?ops_request_misc=%257B%2522request%255Fid%2522%253A%2522166428989816782428613725%2522%252C%2522scm%2522%253A%252220140713.130102334..%2522%257D&request_id=166428989816782428613725&biz_id=0&utm_medium=distribute.pc_search_result.none-task-blog-2~all~top_positive~default-1-123704930-null-null.142^v50^pc_rank_34_2,201^v3^control_2&utm_term=CUDA%E5%AE%89%E8%A3%85&spm=1018.2226.3001.4187https://blog.csdn.net/m0_45447650/article/details/123704930?ops_request_misc=%257B%2522request%255Fid%2522%253A%2522166428989816782428613725%2522%252C%2522scm%2522%253A%252220140713.130102334..%2522%257D&request_id=166428989816782428613725&biz_id=0&utm_medium=distribute.pc_search_result.none-task-blog-2~all~top_positive~default-1-123704930-null-null.142%5Ev50%5Epc_rank_34_2,201%5Ev3%5Econtrol_2&utm_term=CUDA%E5%AE%89%E8%A3%85&spm=1018.2226.3001.4187

4.torch和torchvision安装

（torch和torchvision下载网址：https://download.pytorch.org/whl/torch_stable.html）

(win和linux系统选择，cp36，cu111)

 然后在pycharm虚拟环境的终端pip下载相应whl即可

5.mmcv-full安装（主要是要版本合适）

GitHub - open-mmlab/mmcv at v1.4.7