Research map / Image generation / Autoregressive and latent generation
CAMERAS: Enhanced Resolution and Sanity Preserving Class Activation Mapping for Image Saliency
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 multi-scale accumulation and fusion of activation maps and backpropagated gradients | CAMERAS on PASCAL VOC 2007 test with VGG16 · PASCAL VOC 2007 test, All | Pointing game accuracy: 88.0 -> 86.2 | Hurt |
| Use multi-scale accumulation and fusion of activation maps and backpropagated gradients | CAMERAS on PASCAL VOC 2007 test with VGG16 · PASCAL VOC 2007 test, Diff | Pointing game accuracy: 76.1 -> 76.2 | Helped |
| Use multi-scale accumulation and fusion of activation maps and backpropagated gradients | CAMERAS on PASCAL VOC 2007 test with ResNet50 · PASCAL VOC 2007 test | Pointing game accuracy: 90.4 -> 94.2; 82.3 -> 88.8 | Helped |
| Use multi-scale accumulation and fusion of activation maps and backpropagated gradients | CAMERAS on COCO 2014 validation with VGG16 · COCO 2014 validation | Pointing game accuracy: 54.2 -> 55.4; 49.0 -> 50.7 | Helped |
| Use multi-scale accumulation and fusion of activation maps and backpropagated gradients | CAMERAS on COCO 2014 validation with ResNet50 · COCO 2014 validation | Pointing game accuracy: 57.3 -> 69.9; 52.3 -> 66.4 | Helped |
| Use CAMERAS saliency to focus PGD perturbations on salient regions | PGD on ResNet-50 · ImageNet validation | PGD perturbation norm reduction: 56.5% | Helped |
| Increase the maximum upsampled input size (ablation) | CAMERAS · ImageNet models | Saliency performance | Helped |
| Vary the number of multi-scale steps within N ∈ [5, 10] (ablation) | CAMERAS · ImageNet models | Saliency performance | No clear effect |
Closest papers by meaning
- There and Back Again: Revisiting Backpropagation Saliency Methods (CVPR 2020 · similarity 0.92)
- Sanity Checks for Saliency Maps (NeurIPS 2018 · similarity 0.89)
- Understanding and Visualizing Deep Visual Saliency Models (CVPR 2019 · similarity 0.89)
- DANCE: Enhancing saliency maps using decoys (ICML 2021 · similarity 0.88)
- DD-CAM: Minimal Sufficient Explanations for Vision Models Using Delta Debugging (NeurIPS 2026 · similarity 0.88)
- Great Minds Think Alike: The Universal Convergence Trend of Input Salience (NeurIPS 2024 · 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.