Research map / Architectures and efficiency / RNNs and normalization
Generating Attribution Maps With Disentangled Masked Backpropagation
Adria Ruiz, Antonio Agudo, Francesc Moreno-Noguer
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 disentanglement of positive and negative factors (optimize masks) versus vanilla gradient x input (ablation) | DMBP · ImageNet (5K subset) ResNet50 | IM: .26 -> .50 | Improved |
| add disentanglement versus vanilla gradient x input (ablation) | DMBP · VOC2012 val VGG16 | IM: .36 -> .62 | Improved |
| add negative-term minimization to positive-only loss (ablation) | DMBP · VOC2012 val VGG16 | cIM: .06 -> .18 | Improved |
| add negative-term minimization to positive-only loss (ablation) | DMBP · VOC2012 val VGG16 | IM: .62 -> .51 | Hurt |
| add negative-term minimization to positive-only loss (ablation) | DMBP · VOC2012 val ResNet50 | IM: .58 -> .59 | Improved |
| add nuisance-term minimization (full loss) (ablation) | DMBP · ImageNet (5K subset) ResNet50 | IM: .51 -> .57 | Improved |
| add nuisance-term minimization (full loss) (ablation) | DMBP · ImageNet (5K subset) VGG16 | IM: .37 -> .39 | Improved |
| add nuisance-term minimization (full loss) (ablation) | DMBP · VOC2012 val VGG16 | IM: .51 -> .57 | Hurt |
| DMBP versus GradCam on ImageNet VGG16 | DMBP · ImageNet val VGG16 | IM: 0.48 -> 0.41 | Hurt |
| DMBP versus GradCam average over all settings | DMBP · All settings | Avg of IM/cIM: 0.37 -> 0.42 | Improved |
| DMBP versus SmoothGrad on VOC VGG IM | DMBP · VOC2012 val VGG16 | IM: 0.55 -> 0.57 | Improved |
| DMBP versus GBp in layer reinitialization sanity check | DMBP · VOC VGG16 reinit correlation | rank correlation (lower better): 0.99 -> 0.02 | Improved |
Closest papers by meaning
- Learning Propagation Rules for Attribution Map Generation (ECCV 2020 · similarity 0.89)
- Reliable Evaluation of Attribution Maps in CNNs: A Perturbation-Based Approach (IJCV 2024 · similarity 0.89)
- Towards Better Understanding Attribution Methods (CVPR 2022 · similarity 0.86)
- Benchmarking the Attribution Quality of Vision Models (NeurIPS 2024 · similarity 0.86)
- Pixel-level Certified Explanations via Randomized Smoothing (ICML 2025 · similarity 0.84)
- One Wave To Explain Them All: A Unifying Perspective On Feature Attribution (ICML 2025 · similarity 0.84)
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.
