search

ydhongHIT / DDRNet

The official implementation of "Deep Dual-resolution Networks for Real-time and Accurate Semantic Segmentation of Road Scenes"
MIT License
415 stars 53 forks source link

Error in code or paper? #24

Closed qwert1337 closed 2 years ago

qwert1337 commented 2 years ago

Hi, nice work first of all! I stumbled across the relu activations in the code: x = x + self.down3(self.relu(x_)) x_ = x_ + F.interpolate( self.compression3(self.relu(layers[2])), size=[height_output, width_output], mode='bilinear') https://github.com/ydhongHIT/DDRNet/blob/ba659f9ec358f9ef55dfea00f5e63dae6ad3efd9/segmentation/DDRNet_23_slim.py#L312 In the paper in Fig. 3 there are no activations after blocks, but only after the bilateral fusion. Now I'm wondering what is correct?

ydhongHIT commented 2 years ago

Hi, nice work first of all! I stumbled across the relu activations in the code: x = x + self.down3(self.relu(x_)) x_ = x_ + F.interpolate( self.compression3(self.relu(layers[2])), size=[height_output, width_output], mode='bilinear')

https://github.com/ydhongHIT/DDRNet/blob/ba659f9ec358f9ef55dfea00f5e63dae6ad3efd9/segmentation/DDRNet_23_slim.py#L312

In the paper in Fig. 3 there are no activations after blocks, but only after the bilateral fusion. Now I'm wondering what is correct?

Hi, please follow the code. The pre-trained models work well with the current codes. In fact, one relu won't affect performance a lot.