Tramac/Fast-SCNN-pytorch
A PyTorch Implementation of Fast-SCNN: Fast Semantic Segmentation Network
PythonApache-2.0
Issues
- 0
How can I know the mapping of class and index?
#54 opened by nlc2002 - 0
- 1
the loss decline too slow
#51 opened by DOVEXQ - 1
mIoU and PixelAcc on test set
#18 opened by medhini - 9
Loss does not decrease accordingly
#46 opened by mertmerci - 2
about the miou and pa?
#20 opened by louyanyang - 0
loss too high, mIoU too low
#52 opened by Alohomora0130 - 16
Why doesn't the loss function go down?
#14 opened by LYJvincent - 2
higher speed or forward inference time?
#24 opened by Zeeshan313 - 6
- 10
assert target.dim() == 3 error
#26 opened by midtom - 1
- 0
split train option for segmentation can't do properly and get error all the time
#49 opened by rikkudo - 5
How to train a new model with custom dataset
#35 opened by rose-jinyang - 0
If I want to change the class of the training data, where do I need to change the code?
#48 opened by Sobel-Li - 15
ValueError: Expected more than 1 value per channel when training, got input size torch.Size([1, 32, 1, 1])
#5 opened by 571502680 - 2
Training is too slow
#28 opened by wangq95 - 1
Wrong calculation of IoU
#45 opened by JonasFrey96 - 2
- 6
VOC or COCO dataset model file
#19 opened by shivSD - 16
- 11
Saving / Restoring Weights
#37 opened by mertmerci - 16
Training stuck on the first epoch
#44 opened by mertmerci - 2
ValueError: height and width must be > 0
#41 opened by Linda-L - 2
- 5
Training does not start
#39 opened by mertmerci - 5
Support of testing for any size image
#34 opened by rose-jinyang - 2
Testing on a video
#40 opened by SamSamhuns - 5
predicted mask is whole black
#38 opened by mxy519 - 1
About the demo
#36 opened by shamangary - 3
- 0
作者你好,我想问一下如何训练自己的数据集
#31 opened by Liu6697 - 0
assert not target.requires_grad AssertionError
#30 opened by dronefreak - 1
crop-size issue
#25 opened by zhangyuanxue - 0
RuntimeError:
#29 opened by Sharpiless - 0
Problems while loadeing the dataset.
#27 opened by umgefahren - 0
reducing the noise in the predicted image
#23 opened by sunnysd - 3
COCO dataset : weight tensor should be defined either for all or no classes at
#21 opened by sunnysd - 0
- 0
Why is the accuracy of the document (eval.py) not as good as that of the paper?
#17 opened by wk524629918 - 0
Get horroble result
#16 opened by AntoineGerardeaux - 1
ADE20k dataset supported?
#15 opened by rnditdev - 2
- 0
Why do you generate zero-padded crop image?
#13 opened by priancho - 1
class balancing option is not accessible
#11 opened by priancho - 6
- 0
Multi-gpu setting bug?
#10 opened by priancho - 0
How it can achieve such higher speed?
#8 opened by lqxisok - 1
about train
#6 opened by KaaiLee - 1
_DWconv and _DSconv
#3 opened by TianMingChen