STDC-Seg

STDC-Seg copied to clipboard

当我调用model_stages_trt.py运行的时候，存在size不匹配问题，输入为（1，3，512，512），是否是我未进行正确的操作。 

xinguta001

How can I train with single GPU ?

4

Hello, thank you for sharing the awesome code. I'm using CUDA 11.1, PyTorch 1.8.0 with my single GPU(RTX 3090) and I have tested that evaluation and latency code are working...

CCodie

如何理解trainid=-1

2

作者您好，我在复现您的代码时，发现您的cityscapes_info.json文件中除了trainid=255外，还有一个trainid=-1的分类。然而前面已经有了0-18个trainid，正好为n_classes=19，那么这里trainid=-1的类起到了什么作用呢？

774911840

pretrained model

3

@MichaelFan01 你好!首先感謝您的貢獻，這是一篇很驚艷的作品。 針對paper部分有一點疑惑希望您能幫我解惑 在abstract中有提到 BiseNet backbones borrowed from pretrained tasks, e.g., image classification, may be inefficient for image segmentation due to the deficiency of taskspecific design 您是指Xception39 or ResNet是for Image...

guo0522

No module named 'inplace_abn'

10

训练的时候遇到这个问题，有什么办法解决吗？ python==3.5.6 pytorch==1.1

ZhaoNan1

关于inference time

18

你好，我下载了checkpoint,用STDC1-Seg/model_maxmIOU50.pth，backbone='STDCNet813', scale=0.5进行测试，在RTX2080上运行，跑完cityscapes数据val集的500张图大概需要57s,并未达到实时啊，不知道测试时还需要设置什么参数吗？

gyy1502

TensorRT network must have at least one output and validation failed

3

When I try to run run_latency_stages.py below error occurs. Could you please help me? ...... %657 : Float(1, 19, 64, 128) = onnx::Conv[dilations=[1, 1], group=1, kernel_shape=[1, 1], pads=[0, 0, 0,...

DuyguSerbes

error in running run_latency_stages.py

5

I got the following error when running python ./run_latency_stages.py: [TensorRT] ERROR: Network must have at least one output [TensorRT] ERROR: Network validation failed. Traceback (most recent call last): File "./run_latency_stages.py",...

fjremnav

What's the purpose of the multipication of n_classes to label in evaluation code?

1

https://github.com/MichaelFan01/STDC-Seg/blob/59ff37fbd693b99972c76fcefe97caa14aeb619f/evaluation.py#L56-L59

lx-r

Train on Camvid

2

Please let me know how to train on Camvid, or is there pretrain model on camvid?

odyssey0529