Pytorch-UNet
Pytorch-UNet copied to clipboard
milesial
PyTorch implementation of the U-Net for image semantic segmentation with high quality images
At lines 114-115 in train.py. I believe 'scaler.unscale_(optimizer)' should be added before gradient clipping.
curious but whenever I train a model, it's always using CPU and never gpu. I'm confused why it's not detecting any cuda when i have cuda already installed
Installation fails when running `pip install -r requirements.txt` on Python 3.11.7 and newer. See: https://github.com/nodejs/node-gyp/issues/2869 If it helps, some error logs below. Note that I've tried this on Windows PowerShell....
WSL is the best method of using this package on Windows devices. However, for users who do not have WSL installed yet and want to quickly try out this package,...
Added references to `train_hq.zip` and `train_masks.zip` in the README.
If `download_data.sh` script breaks before the datasets are downloaded, it is better to refer to the exact dataset names so they can be downloaded from Kaggle: `train_hq.zip` `train_masks.zip`
(base) e:\Pytorch-UNet>bash scripts/download_data.sh scripts/download_data.sh: line 17: pip: command not found scripts/download_data.sh: line 19: kaggle: command not found scripts/download_data.sh: line 20: unzip: command not found mv: cannot stat 'train_hq/*': No such...
I set the training batchsize to 4, a total of about 4,500 pieces of training data, according to the speed of training, training rounds take about 10 hours or so,...
 但是输出的结果也是在合理范围内呀  有大佬帮忙看看什么原因吗。谢谢啦
