Zilong Huang
Zilong Huang
The issue may occur when the machine is out of GPU memory. You can reduce the input size and try again
Hi @shrutishrestha Maybe the new branch [pure-python](https://github.com/speedinghzl/CCNet/tree/pure-python) could solve the problem. It's a pure python implementation of CCNet and it does not need to be compiled with CUDA.
@johndpope Thanks for your reply. I recently have updated the [pure-python branch](https://github.com/speedinghzl/CCNet/tree/pure-python).
please try to decrease input size from 769 to 713.
@hmchuong Thanks for your support. It could be better if you could create a pull request, then I will merge the new branch into the repo.
@hmchuong Sorry for the delay. There is a new branch https://github.com/speedinghzl/CCNet/tree/pytorch-1.5. Thanks for the PR.
Hi @Hanqer , we do not carry out experiments on VOC dataset and can not provide training details.
Hi @pcl111, you might want to check which arch your GPU is, and modify run.sh
The 19 elements of arrays represent 19 categories. You are excepted to comprehend mean of IOU. Is your result obtained by carrying out run_local.sh without any modification?