Generating Part-Based Global Explanations Via Correspondence

2025/09/18 by Rathore, Kunal, Tadepalli, Prasad
#Artificial Intelligence (cs.AI) #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences

paper · doi:10.48550/arxiv.2509.15393

Abstract

Deep learning models are notoriously opaque. Existing explanation methods often focus on localized visual explanations for individual images. Concept-based explanations, while offering global insights, require extensive annotations, incurring significant labeling cost. We propose an approach that leverages user-defined part labels from a limited set of images and efficiently transfers them to a larger dataset. This enables the generation of global symbolic explanations by aggregating part-based local explanations, ultimately providing human-understandable explanations for model decisions on a large scale.

Citations

• Visual correspondence-based explanations improve AI robustness and human-AI team accuracy
• PartImageNet: A Large, High-Quality Dataset of Parts
• Image Matching from Handcrafted to Deep Features: A Survey
• Hyperpixel Flow: Semantic Correspondence with Multi-layer Neural Features
• Net2Vec: Quantifying and Explaining how Concepts are Encoded by Filters\n in Deep Neural Networks
• Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)
• Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization
• Densely Connected Convolutional Networks
• Deep Residual Learning for Image Recognition
• Striving for Simplicity: The All Convolutional Net
• Hypercolumns for Object Segmentation and Fine-grained Localization
• Very Deep Convolutional Networks for Large-Scale Image Recognition
• ImageNet Large Scale Visual Recognition Challenge
• ImageNet Large Scale Visual Recognition Challenge
• Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps
• Visualizing and Understanding Convolutional Networks
• SLIC Superpixels Compared to State-of-the-Art Superpixel Methods
• How to Explain Individual Classification Decisions

Related

• ExplainReduce: Generating global explanations from many local explanations
• ConLUX: Concept-Based Local Unified Explanations
• GLANCE: Global to Local Architecture-Neutral Concept-based Explanations
• From attribution maps to human-understandable explanations through Concept Relevance Propagation
• Teaching Meaningful Explanations