Beyond Patches: Mining Interpretable Part-Prototypes for Explainable AI

Keyphrases

• Explainable AI 100%
• AI Systems 100%
• State-of-the-art Techniques 33%
• Deep Learning Model 33%
• Trustworthiness 33%
• Image Classification 33%
• Concept Level 33%
• Image Region 33%
• Challenging Conditions 33%
• Semantic Consistency 33%
• Region Selection 33%
• Classification Benchmark 33%
• Prototype Approach 33%
• Concept Activation Vectors 33%
• Occlusion Type 33%
• Structured Concept 33%
• Interpretable Deep Learning 33%
• Example-based Explanations 33%
• Heatmap 33%
• Concept Mining 33%
• Mining Networks 33%
• Meaningful Image 33%
• AI Alignment 33%
• Gradient-weighted Class Activation Mapping (Grad-CAM) 33%
• Post-hoc Methods 33%
• Rigid Region 33%
• Human Expectation 33%

Computer Science

• Artificial Intelligence 100%
• Explainable Artificial Intelligence 100%
• Interpretability 50%
• Deep Learning Model 25%
• Image Classification 25%
• Semantic Consistency 25%
• Structured Concept 25%