PeliCAM – A deep learning model explanation toolkit built with PyQt
Created by Murali and Sreenath
PeliCAM is a desktop application for visualizing model interpretability.
It supports techniques like:
- CAM (Class Activation Maps)
- LIME (Local Interpretable Model-agnostic Explanations)
- Bounding box annotations
It’s built to help users understand and debug deep learning predictions—visually and interactively.
- Image Loader & Custom PyTorch Model Support
- Layer-wise CAM Selection (Grad-CAM, Layer-CAM)
- Manual & Semi-auto Bounding Box Tool
- LIME Visualizations (positive/negative features)
- Save & Export Visual Outputs
- Tabbed Viewer for easy side-by-side analysis
- Reset + Refresh for clean iteration
- Python
- PyTorch
- PyQt5
- NumPy
- Compare model saliency vs. ground truth regions
- Add interpretability to your PyTorch models
- Fine-tune thresholding and visualization layers
- Save outputs for reports or further analysis
If you're new to Explainable AI (XAI) and want to learn more about:
- What is CAM?
- What is LIME, SHAP, and how they work?
- When to use each method?
Read our detailed XAI Concept Report:
-> XAI Report
Full User Manual available:
-> Read PDF
Download Version v1.0 -> Download. Download Version v1.1 -> Download.
If you like this project, star it ⭐ and share it with your friends!
Let us know how you use PeliCAM!