gasparian / multiclass-semantic-segmentation Goto Github PK

Experiments with UNET/FPN models and cityscapes/kitti datasets [Pytorch]

License: MIT License

Python 100.00%

cityscapes cityscapes-dataset classification deep-learning feature-pyramid-network kitti-dataset multiclass neural-networks pytroch self-driving semantic-segmentation torch unet unet-image-segmentation

multiclass-semantic-segmentation's People

Contributors

Stargazers

Watchers

multiclass-semantic-segmentation's Issues

What is the class weight for 20 classes on cityscape?

I noticed that by defult you set 8 classes for segmentation, and the class weight you set for each of the eight categories is 0.125.
So, if I need to segment on 20 classes, what is the defult-setting class weights? 0.05 for each category?

val single image？

hi,I have a question，how val single image？

File Not Found Error when training

Many thanks for your effort to share this nice project on Github.
However, as I download it and try to run it, I get the below error:
FileNotFoundError: [Errno 2] No such file or directory: "/samsung_drive/semantic_segmentation/UNET_4x_downsize/last_model.pth"

Could you tell me what's going on?
Does it mean you need to provide another pre-trained mode?

Best wishes,
Qi

Recommend Projects