search

chenjun2hao / DDRNet.pytorch

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 cityscapes and camvid, without using inference acceleration and extra data
Other
161 stars 41 forks source link

why `pred.exp()`? #14

Open SauceCat opened 2 years ago

SauceCat commented 2 years ago

Thanks for the nice implementation :cat: :100:, I have a small question:

https://github.com/chenjun2hao/DDRNet.pytorch/blob/bc0e193e87ead839dbc715c48e6bfb059cf21b27/lib/datasets/base_dataset.py#L222 Why pred.exp() in inference? The real scores should be softmax(pred, dim=1) right? Just wonder what is the usage of adding exp() here, is it for visualization?