pytorch-grad-cam and vision_models_visualized
The latter project is a specialized application built using the former more general-purpose Grad-CAM library, making them complements where one provides the core functionality and the other showcases specific use cases and comparisons of vision models.
About pytorch-grad-cam
jacobgil/pytorch-grad-cam
Advanced AI Explainability for computer vision. Support for CNNs, Vision Transformers, Classification, Object detection, Segmentation, Image similarity and more.
Implements 16+ attribution methods ranging from gradient-based approaches (GradCAM, GradCAM++) to perturbation-based techniques (AblationCAM, ScoreCAM) with batched inference for high performance. Built on PyTorch, it supports explainability across diverse architectures including CNNs, Vision Transformers, and multimodal models like CLIP, plus includes built-in metrics and smoothing algorithms to validate and refine explanation quality. Also works with medical imaging, embedding similarity tasks, and provides deep feature factorization for interpretable representation analysis.
About vision_models_visualized
ztsv-av/vision_models_visualized
Project on exploring how different vision models “see” and analyze the images. We compare ConvNeXt, DeiT, and MLP-Mixer using Grad-CAM, attention maps, and saliency, with tools to run inference, analyze results, and visualize model focus regions.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work