mlsanityData to 14 Oct 2026

Test-Time Canonicalization by Foundation Models for Robust Perception

arXiv 2507.10375 ↗First posted 14 Jul 2025cs.CV cs.LGcited by 5

Not on the map. Why?

Abstract (arXiv)

Perception in the real world requires robustness to diverse viewing conditions. Existing approaches often rely on specialized architectures or training with predefined data augmentations, limiting adaptability. Taking inspiration from mental rotation in human vision, we propose FOCAL, a test-time robustness framework that transforms the input into the most typical view. At inference time, FOCAL explores a set of transformed images and chooses the one with the highest likelihood under foundation model priors. This test-time optimization boosts robustness while requiring no retraining or architectural changes. Applied to models like CLIP and SAM, it significantly boosts robustness across a wide range of transformations, including 2D and 3D rotations, contrast and lighting shifts, and day-night changes. We also explore potential applications in active vision. By reframing invariance as a test-time optimization problem, FOCAL offers a general and scalable approach to robustness. Our code is available at: https://github.com/sutkarsh/focal.

Key facts. Cited by 5 papers in the map (data through 14 Oct 2026). Builds on 52 of its 74 references that are in the map. First posted on arXiv 14 Jul 2025.

Builds on

52 of 74 references.

References · 52

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  3. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  4. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
  5. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  6. 2023Segment Anythingcited by 8,234
  7. 20233D Gaussian Splatting for Real-Time Radiance Field Renderingcited by 6,226
  8. 2022LAION-5B: An open large-scale dataset for training next generation image-text modelscited by 4,152
  9. 2016A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networkscited by 2,385
  10. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,986
  11. 2022Objaverse: A Universe of Annotated 3D Objectscited by 1,540
  12. 2018A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attackscited by 1,485
38 more
  1. 2015Spatial Transformer Networkscited by 1,429
  2. 2017Enhancing The Reliability of Out-of-distribution Image Detection in Neural Networkscited by 1,328
  3. 2020Energy-based Out-of-distribution Detectioncited by 1,299
  4. 2019LVIS: A Dataset for Large Vocabulary Instance Segmentationcited by 1,217
  5. 2021Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gaugescited by 1,167
  6. 2016Group Equivariant Convolutional Networkscited by 1,165
  7. 2024Structured 3D Latents for Scalable and Versatile 3D Generationcited by 770
  8. 2016Automatic chemical design using a data-driven continuous representation of moleculescited by 765
  9. 2018A Tutorial on Bayesian Optimizationcited by 744
  10. 2021Common Objects in 3D: Large-Scale Learning and Evaluation of Real-life 3D Category Reconstructioncited by 652
  11. 2015Multi-view Convolutional Neural Networks for 3D Shape Recognitioncited by 565
  12. 2022Open-Vocabulary Semantic Segmentation with Mask-adapted CLIPcited by 565
  13. 2019Your Classifier is Secretly an Energy Based Model and You Should Treat it Like Onecited by 405
  14. 2021Vector Neurons: A General Framework for SO(3)-Equivariant Networkscited by 305
  15. 2018On the Generalization of Equivariance and Convolution in Neural Networks to the Action of Compact Groupscited by 300
  16. 2022Diffusion models as plug-and-play priorscited by 231
  17. 2018A General Theory of Equivariant CNNs on Homogeneous Spacescited by 228
  18. 2020What Should Not Be Contrastive in Contrastive Learningcited by 226
  19. 2020MVTN: Multi-View Transformation Network for 3D Shape Recognitioncited by 140
  20. 2020Sample-Efficient Optimization in the Latent Space of Deep Generative Models via Weighted Retrainingcited by 123
  21. 2022Equivariance with Learned Canonicalization Functionscited by 109
  22. 2022Denoising diffusion models for out-of-distribution detectioncited by 104
  23. 2020Category Level Object Pose Estimation via Neural Analysis-by-Synthesiscited by 100
  24. 2022Local Latent Space Bayesian Optimization over Structured Inputscited by 79
  25. 2025Trading Inference-Time Compute for Adversarial Robustnesscited by 61
  26. 2022ViewFool: Evaluating the Robustness of Visual Recognition to Adversarial Viewpointscited by 48
  27. 2023Equivariant Adaptation of Large Pretrained Modelscited by 41
  28. 2016Fast Fourier Color Constancycited by 37
  29. 2021Grounding inductive biases in natural images:invariance stems from variations in datacited by 27
  30. 2024ImageNet3D: Towards General-Purpose Object-Level 3D Understandingcited by 25
  31. 2023UrbanIR: Large-Scale Urban Scene Inverse Rendering from a Single Videocited by 22
  32. 2023Towards Viewpoint-Invariant Visual Recognition via Adversarial Trainingcited by 19
  33. 2020A Multi-Hypothesis Approach to Color Constancycited by 18
  34. 2023Understanding the Detrimental Class-level Effects of Data Augmentationcited by 13
  35. 2022Do Deep Networks Transfer Invariances Across Classes?cited by 12
  36. 2024A Generative Model of Symmetry Transformationscited by 10
  37. 2024Omniview-Tuning: Boosting Viewpoint Invariance of Vision-Language Pre-training Modelscited by 9
  38. 2016Compensating for Large In-Plane Rotations in Natural Imagescited by 2

50 most cited of 52 All 52 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 2; 2026: 3.

5 papers.

  1. 2025Adaptive Canonicalization with Application to Invariant Anisotropic Geometric Networkscited by 3
  2. 2025Agentic Aerial Cinematography: From Dialogue Cues to Cinematic Trajectoriescited by 2
  3. 2026Inverting Data Transformations via Diffusion Samplingcited by 2
  4. 2026What are the Right Symmetries for Formal Theorem Proving?cited by 1
  5. 2026Zero-Shot Test-Time Canonicalization using Out-of-Distribution Scoringcited by 0