Skip to content

yuan0038/GraphConvNet_extended

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

21 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

GraphConvNet

Hierarchical Convolution and Graph Net for Utilizing Structural Information of Image image

Requirements

Pytorch 1.7.0, timm 0.3.2, torchprofile 0.0.4, apex

Pretrained models on ImageNet

  • GraphConvNet
Model Params (M) FLOPs (B) Top-1 BaiduDisk URL
GraphConvNet-Ti 7.7 1.3 77.1 BaiduDisk URL
GraphConvNet-S 24.5 4.9 82.0 BaiduDisk URL
  • Pyramid GraphConvNet
Model Params (M) FLOPs (B) Top-1 BaiduDisk URL
Pyramid GraphConvNet-Ti 11.4 1.8 80.5 BaiduDisk URL
Pyramid GraphConvNet-S 29.2 4.9 82.4 BaiduDisk URL

Train & Evaluation

see run.sh

Visualization

The visualization code only available to GraphConvNet and ViG

  1. Create a fold named 'ckpt' in './viz_nodes' and download the checkpoints of GraphConvNet-Ti or GraphConvNet-S and put them in './viz_nodes/ckpt'
  2. Open viz_demo.ipnb,and set arguments(arch,)
  3. Run cells

⚠️⚠️⚠️ if you want to visualize ViG, please download the checkpoints I provide here, since I reorganized ViG code and transformed the official checkpoints' state_dict to suit my code.

Demo

  • The first row: gradcam heatmaps of GraphConvNet-Ti in 4th,8th,12th layers.
  • The second row: the patch(node) that has the max gradcam value(most discriminative) and its corresponding neighbors in different layers.
  • The third row: add edges,the pentagram is the most discriminative node.(you can draw edges using tools such as PowerPoint, OmniGraffle..)

Image 1Image 1Image 1 Image 1Image 1Image 1 Image 1Image 1Image 1

Acknowledgement

This repo partially uses code from vig

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published