Research map / Architectures and efficiency / CNN design
Interpretable CNNs for Object Classification
Quanshi Zhang, Xin Wang, Ying Nian Wu, Huilin Zhou, Song Chun Zhu
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 |
|---|---|---|---|
| add interpretable conv-layer with filter loss | VGG-16 · VOC Part (binary) | part interpretability: 0.495 -> 0.770 | Improved |
| add interpretable conv-layer with filter loss | VGG-S · VOC Part (binary) | part interpretability: 0.257 -> 0.390 | Improved |
| add interpretable conv-layer with filter loss | VGG-16 · ILSVRC DET Animal-Part (binary) | location instability: 0.144 -> 0.085 | Improved |
| add interpretable layer vs adding an ordinary conv-layer | VGG-16 · ILSVRC DET Animal-Part (binary) | location instability: 0.121 -> 0.085 | Improved |
| add filter loss (vs mask layer alone) (ablation) | VGG-16 · VOC Part (binary) | semantic purity: 0.370 -> 0.566 | Improved |
| add filter loss (vs mask layer alone) (ablation) | VGG-M · VOC Part (binary) | semantic purity: 0.253 -> 0.465 | Improved |
| add filter loss (vs mask layer alone) (ablation) | VGG-16 · VOC Part (binary) | location instability (first conv-layer): 0.134 -> 0.126 | Improved |
| add filter loss to interpretable CNN (ablation) | VGG-16 · VOC Part (binary) | location instability: 0.094 -> 0.096 | No clear effect |
| add filter loss to interpretable CNN (ablation) | VGG-16 · VOC Part (binary) | classification accuracy: 97.02 -> 95.39 | Hurt |
| use interpretable CNN instead of ordinary CNN | VGG-16 · VOC Part (binary) | classification accuracy: 98.66 -> 95.39 | Hurt |
| use interpretable CNN instead of ordinary CNN | VGG-M · VOC Part (multi-category, softmax) | classification accuracy: 81.93 -> 88.03 | Improved |
| use interpretable CNN instead of ordinary CNN | VGG-16 · VOC Part (multi-category, softmax) | classification accuracy: 89.71 -> 91.60 | Improved |
Closest papers by meaning
- Interpretable Convolutional Neural Networks (CVPR 2018 · similarity 0.95)
- PICNN: A Pathway towards Interpretable Convolutional Neural Networks (AAAI 2024 · similarity 0.90)
- Interpretable Image Classification via Non-parametric Part Prototype Learning (CVPR 2025 · similarity 0.87)
- NeSyFOLD: A Framework for Interpretable Image Classification (AAAI 2024 · similarity 0.86)
- Unveiling the Unseen: Identifiable Clusters in Trained Depthwise Convolutional Kernels (ICLR 2024 · similarity 0.86)
- Interpretable Image Classification with Adaptive Prototype-based Vision Transformers (NeurIPS 2024 · similarity 0.86)
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.
