Junjun2016

[Feature Request] Support APCNet

2

Please support [APCNet ](https://github.com/open-mmlab/mmsegmentation/tree/master/configs/apcnet).

[`y = out[0] if i == 0 else self.fuse_layers[i][0](out[0])`](https://github.com/HRNet/Lite-HRNet/blob/hrnet/models/backbones/litehrnet.py#L633) --> `y=0` [`elif not self.multiscale_output:`](https://github.com/HRNet/Lite-HRNet/blob/hrnet/models/backbones/litehrnet.py#L641) --> `if not self.multiscale_output:` Match HRNet Looking forward to your reply.

Support soft cross entropy loss

2

Support soft cross entropy loss

[Feature] Support HourglassNet

1

## Motivation Support more backbones for semantic segmentation. In this PR, we add HourglassNet, which may train from scratch without ImageNet pre-train. ## Modification - [x] Add HourglassNet backbone and...

[Feature] Support Lite-HRNet

1

## Motivation Support more real-time semantic segmentation models. In this PR, we add Lite-HRNet. ## Modification - [x] Add Lite-HRNet backbone and unitests - [x] Add configs for Lite-HRNet -...

[Feature] Support Res2Net

1

## Motivation Support more backbones for semantic segmentation. In this PR, we add Res2Net. ## Modification - [x] Add Res2Net backbone and unitests - [x] Add configs for Res2Net -...

[Feature] Support ShuffleNetV2

1

## Motivation Support more real-time semantic segmentation models. In this PR, we add ShuffleNetV2. ## Modification - [x] Add ShuffleNetV2 backbone and unitests - [ ] Add configs for ShuffleNetV2...