Research map / Label-efficient and robust learning / Recognition and long-tail
Interpretable Basis Decomposition for Visual Explanation
Bolei Zhou, Yiyou Sun, David Bau, Antonio Torralba
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use deeper backbone for concept classifier features (AlexNet to ResNet50) | Concept classifiers · Broden validation | mAP: 0.625 -> 0.804 | Improved |
| explain with IBD instead of CAM, Grad-CAM and Network Dissection | Interpretable Basis Decomposition · AMT pairwise comparison, four Places365 networks | human preference (5-point Likert) | Improved |
| use IBD explanations to compare ResNet50 against shallower networks on a single image | Interpretable Basis Decomposition · AMT model-trust comparison | human preference | Improved |
| IBD compared with Grad-CAM for judging model trust | Interpretable Basis Decomposition · AMT model-trust comparison | human preference | No clear effect |
| use CAM single heatmap instead of IBD for judging trust | CAM · AMT model-trust comparison | human preference | Hurt |
| longer multi-heatmap explanations versus single heatmap | Interpretable Basis Decomposition · AMT comparison vs CAM and Grad-CAM | human preference | Mixed |
| compare two identical ResNet50 (one mirrored) with IBD or CAM | Interpretable Basis Decomposition · AMT model-trust comparison | human preference | No clear effect |
Closest papers by meaning
- Explanation-based Data Augmentation for Image Classification (NeurIPS 2021 · similarity 0.90)
- IDEA: Unwrapping Visual Black-box Models by Interaction Decomposition (NeurIPS 2026 · similarity 0.89)
- Interpreting Deep Visual Representations via Network Dissection (TPAMI 2018 · similarity 0.89)
- Network Dissection: Quantifying Interpretability of Deep Visual Representations (CVPR 2017 · similarity 0.89)
- Invertible Concept-based Explanations for CNN Models with Non-negative Concept Activation Vectors (AAAI 2021 · similarity 0.89)
- Explainable Neural Networks with Guarantee: A Sparse Estimation Approach (AAAI 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
