https://github.com/jaydeep-work/LIME-SAM

LIME-SAM aims to create an Explainable Artificial Intelligence (XAI) framework for image classification using LIME (Local Interpretable Model-agnostic Explanations) as the base algorithm, with the super-pixel method replaced by Segment Anything by Meta (SAM).
https://github.com/jaydeep-work/LIME-SAM

explainable-ai image-classification lime segment-anything-meta

Last synced: about 2 months ago
JSON representation

LIME-SAM aims to create an Explainable Artificial Intelligence (XAI) framework for image classification using LIME (Local Interpretable Model-agnostic Explanations) as the base algorithm, with the super-pixel method replaced by Segment Anything by Meta (SAM).

• Host: GitHub
• URL: https://github.com/jaydeep-work/LIME-SAM
• Owner: jaydeep-work
• License: apache-2.0
• Created: 2023-04-12T17:13:40.000Z (almost 2 years ago)
• Default Branch: main
• Last Pushed: 2023-04-13T18:58:09.000Z (over 1 year ago)
• Last Synced: 2024-08-03T23:24:04.453Z (5 months ago)
• Topics: explainable-ai, image-classification, lime, segment-anything-meta
• Language: Jupyter Notebook
• Homepage:
• Size: 3.4 MB
• Stars: 32
• Watchers: 8
• Forks: 7
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-segment-anything-extensions - Repo

README

        
# LIME-SAM

LIME-SAM aims to create an Explainable Artificial Intelligence (XAI) framework for image classification using LIME (Local Interpretable Model-agnostic Explanations) as the base algorithm, with the super-pixel method replaced by Segment Anything by Meta (SAM).

# Results

*all the images used in below figure can also be found in 'results' folder.

![](results/LIME-SAM-explaination1.png?raw=true "Title")

# Examples

Keras image classifier  

    



# Roadmap

- [x] Add Keras example

- [ ] Add Pytorch example

- [ ] Try diverse pre-trained model to compare our results with LIME

- [ ] Create python package

# Contact

Jaydeep Dedaniya - [[email protected]]([email protected])



  

    

  

  

    



# Acknowledgments

* [Segment Anything (by Meta)](https://github.com/facebookresearch/segment-anything)

* [lime](https://github.com/marcotcr/lime)