Skip to content

yxheartipp/attr-rollout

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Attr Rollout

Official implementation of Attribution Rollout: A New Way to Interpret Visual Transformer. We provide a jupyter notebook for quickly experience the visualization of our approach

Credits

ViT implementation is based on:

Evaluation experiments is based on:

Reproducing evaluation results

Perturbation Test

CUDA_VISIBLE_DEVICES=1 PYTHONPATH=./:$PYTHONPATH python3 baselines/ViT/generate_visualizations.py --method attr_rollout --imagenet-validation-path path/to/imagenet/'ImageNet 2012 DataSets'
CUDA_VISIBLE_DEVICES=1 PYTHONPATH=./:$PYTHONPATH python3 baselines/ViT/pertubation_eval_from_hdf5.py --method attr_rollout

##Segmentation Results

CUDA_VISIBLE_DEVICES=1 PYTHONPATH=./:$PYTHONPATH python3 baselines/ViT/imagenet_seg_eval.py --method attr_rollout_max --imagenet-seg-path /path/to/gtsegs_ijcv.mat

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors