HRNet / HRNet-Semantic-Segmentation

The OCR approach is rephrased as Segmentation Transformer: https://arxiv.org/abs/1909.11065. This is an official implementation of semantic segmentation for HRNet. https://arxiv.org/abs/1908.07919
Other
3.09k stars 682 forks source link

Is there a version of HRNet called HRNetv2-W28? #266

Open sakethbachu opened 2 years ago

sakethbachu commented 2 years ago

I was going through this paper and I came across the model "HRNetv2-W28". I searched it on the internet but could not find it. Does anyone know about it?

AxMM commented 2 years ago

Hi @sakethbachu! HRNetv2 is the HRNet with the "v2" head for semantic segmentation. The W28 is the width of the Network, is the number of channels of first branch.
In this image you have the HRNet architecture with the "v2" head, which is the last layer. With HRNetV2 representation head, all the low resolutions are bilinear upsampled to high resolution, and concatenate the four representations, followed by a 1 × 1 convolution to mix the four representations.

29-HRNetv2