Research map / Vision-language and multimodal / Visual question answering
Grad-CAM: Visual Explanations From Deep Networks via Gradient-Based Localization
Ramprasaath R. Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, Dhruv Batra
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 Grad-CAM instead of Guided Backpropagation gradients for localization | VGG-16 · ILSVRC-15 val top-1 localization error | top-1 localization error: 61.12 -> 56.51 | Improved |
| use Grad-CAM on the unmodified VGG-16 instead of CAM with a GAP architecture | VGG-16 · ILSVRC-15 val | top-1 classification error: 33.40 -> 30.38 | Improved |
| use Grad-CAM instead of CAM for top-5 localization | VGG-16 · ILSVRC-15 val | top-5 localization error: 45.14 -> 46.41 | Hurt |
| add the ReLU on the weighted combination of feature maps (ablation) | Grad-CAM · ILSVRC-15 localization | localization performance | Improved |
| fuse Grad-CAM with Guided Backpropagation (Guided Grad-CAM) | Guided Backpropagation · PASCAL VOC 2007 human study | human accuracy for class identification: 44.44% -> 61.23% | Improved |
| combine Grad-CAM with Deconvolution | Deconvolution · PASCAL VOC 2007 human study | human accuracy for class identification: 53.33% -> 61.23% | Improved |
| use Guided Grad-CAM instead of Guided Backpropagation to compare VGG and AlexNet | trust evaluation · PASCAL VOC 2007 human study | relative reliability score of VGG: 1.00 -> 1.27 | Improved |
| use Grad-CAM instead of Guided Backpropagation, c-MWP or CAM for faithfulness to occlusion | VGG-16 · PASCAL VOC 2007 val | rank correlation with occlusion map: 0.168 -> 0.254 | Improved |
| remove gender bias by adding male nurses and female doctors to training data | VGG-16 doctor vs nurse classifier · doctor vs nurse test set | accuracy: 82% -> 90% | Improved |
| use Grad-CAM instead of Guided Backpropagation for localizing captions | neuraltalk2 captioning · DenseCap region boxes | inside/outside activation ratio: 2.32 -> 3.27 | Improved |
| add Grad-CAM to Guided Backpropagation (Guided Grad-CAM) for caption localization | neuraltalk2 captioning · DenseCap region boxes | inside/outside activation ratio: 3.27 -> 6.38 | Improved |
| use Grad-CAM instead of Guided Backpropagation for VQA faithfulness | VQA LSTM-CNN model · VQA | rank correlation with occlusion map: 0.42 -> 0.60 | Improved |
Later papers that compared with this method
10 later accepted papers report GradCAM as a baseline in their results tables; 9 of them report a better number on most of the results they share with it.
- Rethinking Visual Attribution for Chest X-ray Reasoning in Large Vision Language Models (NeurIPS 2026 · better on 5 of 5 shared results)
- DEX-AR: A Dynamic Explainability Method for Autoregressive Vision-Language Models (ECCV 2026 · better on 3 of 3 shared results)
- Disentangling Hallucinations: Orthogonal Semantic Projection for Robust Interpretability (ECCV 2026 · better on 3 of 3 shared results)
- Grad-ECLIP: Gradient-Based Visual and Textual Explanations for CLIP (TPAMI 2026 · better on 1 of 1 shared result)
- BRACTIVE: A Brain Activation Approach to Human Visual Brain Learning (TPAMI 2025 · better on 5 of 5 shared results)
- Token Activation Map to Visually Explain Multimodal LLMs (ICCV 2025 · better on 3 of 3 shared results)
- Decomposing and Interpreting Image Representations via Text in ViTs Beyond CLIP (NeurIPS 2024 · better on 12 of 12 shared results)
- Interpreting CLIP's Image Representation via Text-Based Decomposition (ICLR 2024 · better on 3 of 3 shared results)
2 more not listed.
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DD-CAM: Minimal Sufficient Explanations for Vision Models Using Delta Debugging (NeurIPS 2026 · similarity 0.89)
- U-CAM: Visual Explanation Using Uncertainty Based Class Activation Maps (ICCV 2019 · similarity 0.88)
- Empowering CAM-Based Methods with Capability to Generate Fine-Grained and High-Faithfulness Explanations (AAAI 2024 · similarity 0.87)
- Finer-CAM: Spotting the Difference Reveals Finer Details for Visual Explanation (CVPR 2025 · similarity 0.87)
- Attention Guided CAM: Visual Explanations of Vision Transformer Guided by Self-Attention (AAAI 2024 · similarity 0.86)
- Improved Visual Grounding through Self-Consistent Explanations (CVPR 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.
