mlsanityData to 7 Oct 2026

Axiom-based Grad-CAM: Towards Accurate Visualization and Explanation of CNNs

arXiv 2008.02312 ↗First posted 5 Aug 2020cs.CV cs.AI cs.LGcited by 151

Not on the map. Why?

Abstract (arXiv)

To have a better understanding and usage of Convolution Neural Networks (CNNs), the visualization and interpretation of CNNs has attracted increasing attention in recent years. In particular, several Class Activation Mapping (CAM) methods have been proposed to discover the connection between CNN's decision and image regions. In spite of the reasonable visualization, lack of clear and sufficient theoretical support is the main limitation of these methods. In this paper, we introduce two axioms -- Conservation and Sensitivity -- to the visualization paradigm of the CAM methods. Meanwhile, a dedicated Axiom-based Grad-CAM (XGrad-CAM) is proposed to satisfy these axioms as much as possible. Experiments demonstrate that XGrad-CAM is an enhanced version of Grad-CAM in terms of conservation and sensitivity. It is able to achieve better visualization performance than Grad-CAM, while also be class-discriminative and easy-to-implement compared with Grad-CAM++ and Ablation-CAM. The code is available at https://github.com/Fu0511/XGrad-CAM.

Key facts. Cited by 151 papers in the map (data through 7 Oct 2026). First posted on arXiv 5 Aug 2020.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 14; 2022: 17; 2023: 34; 2024: 43; 2025: 31; 2026: 12.

151 papers.

Sort:

By area:

  1. 2022Learning Affordance Grounding from Exocentric Imagescited by 82
  2. 2021Towards Better Explanations of Class Activation Mappingcited by 47
  3. 2024Enhancing Descriptive Image Quality Assessment with A Large-scale Multi-modal Datasetcited by 38
  4. 2022Character-Centric Story Visualization via Visual Planning and Token Alignmentcited by 32
  5. 2021F-CAM: Full Resolution Class Activation Maps via Guided Parametric Upscalingcited by 26
  6. 2021Revisiting The Evaluation of Class Activation Mapping for Explainability: A Novel Metric and Experimental Analysiscited by 25
  7. 2021Learning Visual Affordance Grounding from Demonstration Videoscited by 25
  8. 2023Opti-CAM: Optimizing saliency maps for interpretabilitycited by 25
  9. 2022Discovering Transferable Forensic Features for CNN-generated Images Detectioncited by 24
  10. 2023Visual Explanations via Iterated Integrated Attributionscited by 18
  11. 2021Integrated Grad-CAM: Sensitivity-Aware Visual Explanation of Deep Convolutional Networks via Integrated Gradient-Based Scoringcited by 16
  12. 2021Keep CALM and Improve Visual Feature Attributioncited by 15
38 more
  1. 2025Finer-CAM: Spotting the Difference Reveals Finer Details for Visual Explanationcited by 14
  2. 2022Discriminative Sampling of Proposals in Self-Supervised Transformers for Weakly Supervised Object Localizationcited by 13
  3. 2022Explainable Model-Agnostic Similarity and Confidence in Face Verificationcited by 13
  4. 2022TCAM: Temporal Class Activation Maps for Object Localization in Weakly-Labeled Unconstrained Videoscited by 12
  5. 2025Dynamic Accumulated Attention Map for Interpreting Evolution of Decision-Making in Vision Transformercited by 12
  6. 2022Re-Attention Transformer for Weakly Supervised Object Localizationcited by 11
  7. 2023Learning Pairwise Interaction for Generalizable DeepFake Detectioncited by 11
  8. 2024Guided Interpretable Facial Expression Recognition via Spatial Action Unit Cuescited by 11
  9. 2023Towards Trust of Explainable AI in Thyroid Nodule Diagnosiscited by 10
  10. 2023SketchXAI: A First Look at Explainability for Human Sketchescited by 10
  11. 2023Keep the Faith: Faithful Explanations in Convolutional Neural Networks for Case-Based Reasoningcited by 10
  12. 2025Winsor-CAM: Human-Tunable Visual Explanations from Deep Networks via Layer-Wise Winsorizationcited by 10
  13. 2023Deep Integrated Explanationscited by 9
  14. 2022Negative Evidence Matters in Interpretable Histology Image Classificationcited by 8
  15. 2023Explain To Me: Salience-Based Explainability for Synthetic Face Detection Modelscited by 8
  16. 2023DiPS: Discriminative Pseudo-Label Sampling with Self-Supervised Transformers for Weakly Supervised Object Localizationcited by 8
  17. 2024Efficient Bias Mitigation Without Privileged Informationcited by 8
  18. 2024Benchmarking XAI Explanations with Human-Aligned Evaluationscited by 8
  19. 2022FD-CAM: Improving Faithfulness and Discriminability of Visual Explanation for CNNscited by 7
  20. 2023CoLo-CAM: Class Activation Mapping for Object Co-Localization in Weakly-Labeled Unconstrained Videoscited by 7
  21. 2023G-CAME: Gaussian-Class Activation Mapping Explainer for Object Detectorscited by 7
  22. 2024TexTile: A Differentiable Metric for Texture Tileabilitycited by 7
  23. 2024XEdgeAI: A Human-centered Industrial Inspection Framework with Data-centric Explainable Edge AI Approachcited by 7
  24. 2024Benchmarking the Attribution Quality of Vision Modelscited by 7
  25. 2024Benchmarking and Learning Multi-Dimensional Quality Evaluator for Text-to-3D Generationcited by 7
  26. 2021Axiomatic Explanations for Visual Search, Retrieval, and Similarity Learningcited by 6
  27. 2021Towards explainable artificial intelligence (XAI) for early anticipation of traffic accidentscited by 6
  28. 2023How Will It Drape Like? Capturing Fabric Mechanics from Depth Imagescited by 6
  29. 2024Respect the model: Fine-grained and Robust Explanation with Sharing Ratio Decompositioncited by 6
  30. 2024Efficient and Concise Explanations for Object Detection with Gaussian-Class Activation Mapping Explainercited by 6
  31. 2024Improving Accuracy-robustness Trade-off via Pixel Reweighted Adversarial Trainingcited by 6
  32. 2025Generalizable and Explainable Deep Learning for Medical Image Computing: An Overviewcited by 6
  33. 2025FOCUS: Fused Observation of Channels for Unveiling Spectracited by 6
  34. 2023Empowering CAM-Based Methods with Capability to Generate Fine-Grained and High-Faithfulness Explanationscited by 5
  35. 2024Explaining the Implicit Neural Canvas: Connecting Pixels to Neurons by Tracing their Contributionscited by 5
  36. 2024Feature Accentuation: Revealing 'What' Features Respond to in Natural Imagescited by 5
  37. 2024Fully Exploiting Every Real Sample: SuperPixel Sample Gradient Model Stealingcited by 5
  38. 2024Disentangling spatio-temporal knowledge for weakly supervised object detection and segmentation in surgical videocited by 5

50 most cited of 151 All 151 (JSON)