This repository is a pytorch version of sd-maskrcnn and maskrcnn. The offical sd-maskrcnn code is written by tensorflow.(refer to https://github.com/BerkeleyAutomation/sd-maskrcnn). And the maskrcnn backbone is mostly adapted from https://github.com/darolt/mask_rcnn. I make some modification as below:
The code can run with multi-gpus and I tested on
./datasets/wisdom/wisdomConfig.yml
(change the dataset path)CUDA_VISIBLE_DEVICES=X,X python main/trian.py
./log
directory.tensorboardX
package and you can monitor the training procedure in browser../datasets/wisdom/widomInference.yml
CUDA_VISIBLE_DEVICES=0 python main/inference.py