https://github.com/zer0int/clip-vit-visualization

What do CLIP Vision Transformers learn? Feature Visualization can show you!
https://github.com/zer0int/clip-vit-visualization

activation clip feature neuron vision-transformer visualization vit xai

Last synced: 3 months ago
JSON representation

What do CLIP Vision Transformers learn? Feature Visualization can show you!

• Host: GitHub
• URL: https://github.com/zer0int/clip-vit-visualization
• Owner: zer0int
• Created: 2024-04-02T23:17:07.000Z (over 1 year ago)
• Default Branch: CLIP-vision
• Last Pushed: 2024-08-29T08:10:59.000Z (about 1 year ago)
• Last Synced: 2025-04-28T19:51:38.917Z (5 months ago)
• Topics: activation, clip, feature, neuron, vision-transformer, visualization, vit, xai
• Language: Python
• Homepage:
• Size: 26.4 KB
• Stars: 9
• Watchers: 1
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
![generate-some-vit-viz](https://github.com/zer0int/CLIP-ViT-visualization/assets/132047210/6399dc8d-9990-460c-866c-81ef0a25db05)

# CLIP-ViT-Visualization

Standing on the shoulders of giants:

- Based on [github.com/hamidkazemi22/vit-visualization](https://github.com/hamidkazemi22/vit-visualization)

- This repo is a lightweight, CLIP ViT feature visualization only, implementation thereof.

- Contains ALL (!) CLIP models + some speed optimization. Requires CUDA.

- Uses [OpenAI / CLIP](https://github.com/openai/CLIP)

## Info

- Check / install `requirements.txt`

- Check the comments in "run_visualization.py" -> Easy lightweight configuration

- Configure: model, optimizer, single-feature vs. multi-feature/layer, intermediate steps, ...

- From the console, use "python run_visualization.py"

- Saves output in neatly organized and named way (subfolders)

- 29-AUG-204: Added script for deterministic results.

## Warning about CLIP feature visualizations

- CLIP has been trained on "pretty much the entire internet". 

- One neuron gets you a cat, a rose; another may encode explicit / sensitive / offensive / violent content. 

- Use responsibly / at your own discretion.

- For more information, refer to the [CLIP Model Card](https://github.com/openai/CLIP/blob/main/model-card.md).