Skip to content

visinf/what-is-missing

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

What is Missing? Explaining Neurons Activated by Absent Concepts

License Framework

R. Hesse, S. Schaub-Meyer, J. Hesse, B. Schiele, and S. Roth. What is Missing? Explaining Neurons Activated by Absent Concepts. ICML, 2026.

ArXiv | Poster

Environment

conda env create -f environment.yml

conda activate what-is-missing

Reichardt detector experiment

cd reichardt_detector and run python reichardt_detector.py

The results will be stored in /visualization.

Toy experiment

cd toy_example and run python toy_example.py

The results will be stored in /visualization.

ImageNet experiment

cd imagenet run, e.g.,

python image_quantitative.py --data_dir /fastdata/rhesse/datasets/imagenet --model resnet50 --model_layer model.layer4[2].conv3 --batch_size 256 --patch_size 48 --patch_stride 16 --nr_patches 8 --seed 0

ISIC experiment

Download the dataset

cd isic and set the DOWNLOAD_DIR and API_TOKEN in isic_download.py

run python isic_download.py

Disclaimer: the ISIC API changed over the course of the project, and I did not manage to recover exactly the same images. While the overall conclusions should remain the same, it is possible that the downloaded images do not exactly match the images used in the paper.

Running the experiment

run python train.py

You can set:

The debiasing mode - MODE = ['default', 'presence_debias', 'presence_absence_debias']

The model - MODEL = ['xresnet50', 'vit_b_16']

and other parameters (remember to adjust the STORE_DIR for the different models)

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages