mlsanityData to 14 Oct 2026

Understanding Deep Image Representations by Inverting Them

arXiv 1412.0035 ↗First posted 26 Nov 2014cs.CVcited by 403

Not on the map. Why?

Abstract (arXiv)

Image representations, from SIFT and Bag of Visual Words to Convolutional Neural Networks (CNNs), are a crucial component of almost any image understanding system. Nevertheless, our understanding of them remains limited. In this paper we conduct a direct analysis of the visual information contained in representations by asking the following question: given an encoding of an image, to which extent is it possible to reconstruct the image itself? To answer this question we contribute a general framework to invert representations. We show that this method can invert representations such as HOG and SIFT more accurately than recent alternatives while being applicable to CNNs too. We then use this technique to study the inverse of recent state-of-the-art CNN image representations for the first time. Among our findings, we show that several layers in CNNs retain photographically accurate information about the image, with different degrees of geometric and photometric invariance.

Key facts. Cited by 403 papers in the map (data through 14 Oct 2026). First posted on arXiv 26 Nov 2014.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2021: 78; 2022: 91; 2023: 84; 2024: 65; 2025: 50; 2026: 35.

403 papers.

Sort:

By area:

  1. 2021PARE: Part Attention Regressor for 3D Human Body Estimationcited by 337
  2. 2023DreamSim: Learning New Dimensions of Human Visual Similarity using Synthetic Datacited by 337
  3. 2021See through Gradients: Image Batch Recovery via GradInversioncited by 326
  4. 2022Splicing ViT Features for Semantic Appearance Transfercited by 209
  5. 2023Text-To-4D Dynamic Scene Generationcited by 205
  6. 2023Interpreting CLIP's Image Representation via Text-Based Decompositioncited by 196
  7. 2023Text Embeddings Reveal (Almost) As Much As Textcited by 192
  8. 2022Fake it till you make it: Learning transferable representations from synthetic ImageNet clonescited by 172
  9. 2022From Attribution Maps to Human-Understandable Explanations through Concept Relevance Propagationcited by 141
  10. 2023Squeeze, Recover and Relabel: Dataset Condensation at ImageNet Scale From A New Perspectivecited by 138
  11. 2022Natural Language Descriptions of Deep Visual Featurescited by 137
  12. 2022Reconstructing Training Data from Trained Neural Networkscited by 133
38 more
  1. 2023Similarity of Neural Network Models: A Survey of Functional and Representational Measurescited by 131
  2. 2022Toward Transparent AI: A Survey on Interpreting the Inner Structures of Deep Neural Networkscited by 129
  3. 2023Space-Time Diffusion Features for Zero-Shot Text-Driven Motion Transfercited by 124
  4. 2021Interpretable Deep Learning: Interpretation, Interpretability, Trustworthiness, and Beyondcited by 118
  5. 2021Salient ImageNet: How to discover spurious features in Deep Learning?cited by 114
  6. 2021Guided Integrated Gradients: An Adaptive Path Method for Removing Noisecited by 102
  7. 2023Wider and Deeper LLM Networks are Fairer LLM Evaluatorscited by 94
  8. 2022What do Vision Transformers Learn? A Visual Explorationcited by 76
  9. 2022Variational Model Inversion Attackscited by 75
  10. 2021FCA: Learning a 3D Full-coverage Vehicle Camouflage for Multi-view Physical Adversarial Attackcited by 74
  11. 2022Auditing Privacy Defenses in Federated Learning via Generative Gradient Leakagecited by 72
  12. 2022On scientific understanding with artificial intelligencecited by 70
  13. 2023Language Model Inversioncited by 68
  14. 2022Physical Adversarial Attack meets Computer Vision: A Decade Surveycited by 67
  15. 2024SelfIE: Self-Interpretation of Large Language Model Embeddingscited by 67
  16. 2021Contrastive Model Inversion for Data-Free Knowledge Distillationcited by 66
  17. 2021High Fidelity Visualization of What Your Self-Supervised Representation Knows Aboutcited by 66
  18. 2024A Multimodal Automated Interpretability Agentcited by 63
  19. 2025Universal Sparse Autoencoders: Interpretable Cross-Model Concept Alignmentcited by 62
  20. 2022Visual Recognition with Deep Nearest Centroidscited by 59
  21. 2023CodeLMSec Benchmark: Systematically Evaluating and Finding Security Vulnerabilities in Black-Box Code Language Modelscited by 55
  22. 2024Learning by Reconstruction Produces Uninformative Features For Perceptioncited by 55
  23. 2024ConceptPrune: Concept Editing in Diffusion Models via Skilled Neuron Pruningcited by 55
  24. 2022GradViT: Gradient Inversion of Vision Transformerscited by 53
  25. 2024Explainable and Interpretable Multimodal Large Language Models: A Comprehensive Surveycited by 52
  26. 2022Orchestra: Unsupervised Federated Learning via Globally Consistent Clusteringcited by 45
  27. 2021Categorical Foundations of Gradient-Based Learningcited by 44
  28. 2023Recent Advances on Federated Learning: A Systematic Surveycited by 41
  29. 2024Extracting Prompts by Inverting LLM Outputscited by 41
  30. 2021Towards General Deep Leakage in Federated Learningcited by 40
  31. 2022A Survey on Gradient Inversion: Attacks, Defenses and Future Directionscited by 40
  32. 2023DAP: A Dynamic Adversarial Patch for Evading Person Detectorscited by 40
  33. 2021Biometrics: Trust, but Verifycited by 39
  34. 2022On the Explainability of Natural Language Processing Deep Modelscited by 38
  35. 2022Compressed-VFL: Communication-Efficient Learning with Vertically Partitioned Datacited by 36
  36. 2021One-Class Classification: A Surveycited by 35
  37. 2021A Little Robustness Goes a Long Way: Leveraging Robust Features for Targeted Transfer Attackscited by 34
  38. 2022Explaining Deepfake Detection by Analysing Image Matchingcited by 34

50 most cited of 403 All 403 (JSON)