Zilong Huang
Zilong Huang
Maybe you can reduce the batch size to fit your GPU memory.
@ymli39 You need to try train.py rather than train_msc.py.
Thanks for your interest in our work. The seed cues which stored in the pickle file is very important. In every training iteration, the current supervision grows starting from the...
The label map can still dynamic update because of the constantly changed segmentation networks during training. We use the CAM to generate the initial seed (cues pickle file) offline. Afterward,...
Hi @yaoqi-zd @lianxxx, I used ImageNet pre-trained weights for initialization. You can find the pre-trained weights from https://github.com/KaimingHe/deep-residual-networks.
This [repo](https://github.com/ascust/SEC-MXNet) contains the code to generate the foreground localization. Hope it helps.
Only the images are needed in the training step. I can not figure out why you get such performance. It may be helpful if you can upload the training log...
The output of run.sh.
I have no idea after reading the log file. Can you evaluate the model generated by the first step (before retraining step)?
The result of step1 (before retaining) is significantly lower than the result reported in the paper. It's close to the result of the model without DSRG. Maybe you could remove...