DDRNet.pytorch
DDRNet.pytorch copied to clipboard
chenjun2hao
This is the unofficial code of Deep Dual-resolution Networks for Real-time and Accurate Semantic Segmentation of Road Scenes. which achieve state-of-the-art trade-off between accuracy and speed on ci...
https://github.com/chenjun2hao/DDRNet.pytorch/blob/bc0e193e87ead839dbc715c48e6bfb059cf21b27/lib/utils/utils.py#L35 "and" should be "or"
大佬您好，在看您代码的时候发现config文件中BALANCE_WEIGHTS设置的是[1,0.4]，为什么将seghead出来的loss的权重设置成1，而最终输出的权重的loss设置为0.4呢，期待您的回复，十分感谢！
i want to train only two classes. one class + background , can you pls. tell me how i modify the code ? thanks!
I changed TEST_SET in . Yaml and then ran eval.py ,but got an error. ValueError: not enough values to unpack (expected at least 4, got 3)
Why i not found in ddrnet_39.py about Bottleneck_last，SPP_super, So now only use Bottleneck and PPAM replace?
why only Support SGD optimizer, why not adam, is there any Evidence-based for the chose of optimizer?
RuntimeError: value cannot be converted to type float without overflow: (3.32193e-07,-1.07936e-07)
how to solve this running error?
There are calls to some non-existent network blocks in [ddrnet_39.py](https://github.com/chenjun2hao/DDRNet.pytorch/blob/main/lib/models/ddrnet_39.py). More specifically: * `Bottleneck_last` in line 58 * `SPP_super` in line 271 Judging by the [ddrnet_23_slim.py](https://github.com/chenjun2hao/DDRNet.pytorch/blob/main/lib/models/ddrnet_23_slim.py) that has no such...
Metadata
Owner
Metadata
This is the unofficial code of Deep Dual-resolution Networks for Real-time and Accurate Semantic Segmentation of Road Scenes. which achieve state-of-the-art trade-off between accuracy and speed on ci...