Hang Zhang

Results 10 issues of Hang Zhang

Using this issue to keep track of the **state-of-the-art** achieved using **ResNeSt** models: ## Instance Segmentation [![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/resnest-split-attention-networks/instance-segmentation-on-coco-minival)](https://paperswithcode.com/sota/instance-segmentation-on-coco-minival?p=resnest-split-attention-networks) [![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/resnest-split-attention-networks/instance-segmentation-on-coco)](https://paperswithcode.com/sota/instance-segmentation-on-coco?p=resnest-split-attention-networks) [![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/resnest-split-attention-networks/object-detection-on-coco)](https://paperswithcode.com/sota/object-detection-on-coco?p=resnest-split-attention-networks) [![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/resnest-split-attention-networks/object-detection-on-coco-minival)](https://paperswithcode.com/sota/object-detection-on-coco-minival?p=resnest-split-attention-networks) Method Backbone bbox mask Cascade R-CNN ResNeSt-101 (ours) 48.30 41.56...

good first issue

Due to hight downloading volume violate the Wasabi cloud rule, please manually download the models for now. https://drive.google.com/file/d/10GXXl9ekgD3-npa7LiXZpQg6FI1bwpky/view?usp=sharing https://drive.google.com/file/d/1e-Z88a1c14Cwwn02CiBzQ2kFwO6OozHd/view?usp=sharing https://drive.google.com/file/d/1uSSvy4V7ALjousc7Tqy3tlGH7nLUijhG/view?usp=sharing https://drive.google.com/file/d/1dwMhRiuz-E7je-gK0mEFrIXp33MU2_FP/view?usp=sharing

As the major training pipeline has been migrate to PyTorch, we will move the MXNet/Gluon related code to a separate repo. MXNet users can still use the old version of...

Just in case someone want to use the Split-Attention Module. The module is provided here: ```python import torch import torch.nn as nn from torch.nn import functional as F class rSoftMax(nn.Module):...

First, thanks for the awesome work of re-implementing RegNet. I am having difficulty of reproducing the results for `RegNetY-0.4GF`. The configurations are taken from the [original repo](https://github.com/facebookresearch/pycls/blob/master/configs/dds_baselines/regnety/RegNetY-400MF_dds_8gpu.yaml): ``` group_width =...

![image](https://user-images.githubusercontent.com/8041160/94355924-6dcba600-003d-11eb-8578-79af465431dc.png) ![image](https://user-images.githubusercontent.com/8041160/94355926-72905a00-003d-11eb-877f-87a3d5871e27.png) [SyncBN Equations.pdf](https://github.com/zhanghang1989/PyTorch-Encoding/files/5287716/SyncBN.Equations.pdf) [SyncBN Implementation.docx](https://github.com/zhanghang1989/PyTorch-Encoding/files/5287717/SyncBN.Implementation.docx)

It seems only use the package I have already installed locally. It does not consider the package changes between different branches. Therefore, I can only build the master branch successfully....