mlsanityData to 14 Oct 2026

Do It Yourself: Learning Semantic Correspondence from Pseudo-Labels

arXiv 2506.05312 ↗First posted 5 Jun 2025cs.CVcited by 13

Not on the map. Why?

Abstract (arXiv)

Finding correspondences between semantically similar points across images and object instances is one of the everlasting challenges in computer vision. While large pre-trained vision models have recently been demonstrated as effective priors for semantic matching, they still suffer from ambiguities for symmetric objects or repeated object parts. We propose improving semantic correspondence estimation through 3D-aware pseudo-labeling. Specifically, we train an adapter to refine off-the-shelf features using pseudo-labels obtained via 3D-aware chaining, filtering wrong labels through relaxed cyclic consistency, and 3D spherical prototype mapping constraints. While reducing the need for dataset-specific annotations compared to prior work, we establish a new state-of-the-art on SPair-71k, achieving an absolute gain of over 4% and of over 7% compared to methods with similar supervision requirements. The generality of our proposed approach simplifies the extension of training to other data sources, which we demonstrate in our experiments.

Key facts. Cited by 13 papers in the map (data through 14 Oct 2026). Builds on 58 of its 69 references that are in the map. First posted on arXiv 5 Jun 2025.

Builds on

58 of 69 references.

References · 58

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
  3. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,974
  4. 2021Emerging Properties in Self-Supervised Vision Transformerscited by 6,968
  5. 2024SAM 2: Segment Anything in Images and Videoscited by 2,863
  6. 2019Self-training with Noisy Student improves ImageNet classificationcited by 1,311
  7. 2015ORB-SLAM: a Versatile and Accurate Monocular SLAM Systemcited by 988
  8. 2024Structured 3D Latents for Scalable and Versatile 3D Generationcited by 770
  9. 2019Self-labelling via simultaneous clustering and representation learningcited by 528
  10. 2023Emergent Correspondence from Image Diffusioncited by 446
  11. 2021Deep ViT Features as Dense Visual Descriptorscited by 330
  12. 2016LIFT: Learned Invariant Feature Transformcited by 306
38 more
  1. 2023A Tale of Two Features: Stable Diffusion Complements DINO for Zero-Shot Semantic Correspondencecited by 301
  2. 2022TAP-Vid: A Benchmark for Tracking Any Point in a Videocited by 263
  3. 2022Unsupervised Semantic Segmentation by Distilling Feature Correspondencescited by 241
  4. 2023Diffusion Hyperfeatures: Searching Through Time and Space for Semantic Correspondencecited by 214
  5. 2019SPair-71k: A Large-scale Benchmark for Semantic Correspondencecited by 146
  6. 2017Convolutional neural network architecture for geometric matchingcited by 138
  7. 2022AiATrack: Attention in Attention for Transformer Visual Trackingcited by 137
  8. 2021AP-10K: A Benchmark for Animal Pose Estimation in the Wildcited by 123
  9. 2023Unsupervised Semantic Correspondence Using Stable Diffusioncited by 120
  10. 2016Learning Dense Correspondence via 3D-guided Cycle Consistencycited by 91
  11. 2023Telling Left from Right: Identifying Geometry-Aware Semantic Correspondencecited by 86
  12. 2017End-to-end weakly-supervised semantic alignmentcited by 60
  13. 2021DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervisioncited by 59
  14. 2024COVE: Unleashing the Diffusion Feature Correspondence for Consistent Video Editingcited by 59
  15. 2017SCNet: Learning Semantic Correspondencecited by 48
  16. 2024Orient Anything: Learning Robust Object Orientation Estimation from Rendering 3D Modelscited by 45
  17. 2024CleanDIFT: Diffusion Features without Noisecited by 44
  18. 2021Warp Consistency for Unsupervised Learning of Dense Correspondencescited by 42
  19. 2022RNNPose: Recurrent 6-DoF Object Pose Refinement with Robust Correspondence Field Estimation and Pose Optimizationcited by 38
  20. 2023SD4Match: Learning to Prompt Stable Diffusion Model for Semantic Matchingcited by 38
  21. 2023Improving Semantic Correspondence with Viewpoint-Guided Spherical Mapscited by 34
  22. 2022Learning Semantic Correspondence with Sparse Annotationscited by 31
  23. 2023Neural Congealing: Aligning Images to a Joint Semantic Atlascited by 31
  24. 2022Probabilistic Warp Consistency for Weakly-Supervised Semantic Correspondencescited by 30
  25. 2023ASIC: Aligning Sparse in-the-wild Image Collectionscited by 27
  26. 2024LiFT: A Surprisingly Simple Lightweight Feature Transform for Dense ViT Descriptorscited by 27
  27. 2024ImageNet3D: Towards General-Purpose Object-Level 3D Understandingcited by 25
  28. 2022Self-Supervised Geometric Correspondence for Category-Level 6D Object Pose Estimation in the Wildcited by 22
  29. 2021To The Point: Correspondence-driven monocular 3D category reconstructioncited by 21
  30. 2022Neural Correspondence Field for Object Pose Estimationcited by 21
  31. 2019Semantic Attribute Matching Networkscited by 20
  32. 2019Show, Match and Segment: Joint Weakly Supervised Learning of Semantic Matching and Object Co-segmentationcited by 20
  33. 2021The Functional Correspondence Problemcited by 20
  34. 2023NCP: Neural Correspondence Prior for Effective Unsupervised Shape Matchingcited by 18
  35. 2023Back to 3D: Few-Shot 3D Keypoint Detection with Back-Projected 2D Featurescited by 17
  36. 2024Distillation of Diffusion Features for Semantic Correspondencecited by 17
  37. 2022Semi-Supervised Learning of Semantic Correspondence with Pseudo-Labelscited by 16
  38. 2024A Review of Pseudo-Labeling for Computer Visioncited by 16

50 most cited of 58 All 58 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 3; 2026: 10.

13 papers.

Sort:

By area:

  1. 2025Attention (as Discrete-Time Markov) Chainscited by 11
  2. 2025GECO: Geometrically Consistent Embedding with Lightspeed Inferencecited by 6
  3. 2025Towards Geometry-Grounded Dense Semantic Matching with VGGT Priorscited by 3
  4. 2026MARCO: Navigating the Unseen Space of Semantic Correspondencecited by 3
  5. 2026Shape-of-You: Fused Gromov-Wasserstein Optimal Transport for Semantic Correspondence in-the-Wildcited by 1
  6. 2026Geometry Matters: 3D Foundation Priors for Learning Semantic Correspondencecited by 1
  7. 2026PASR: Pose-Aware 3D Shape Retrieval from Occluded Single Viewscited by 0
  8. 2026SEMAGIC: Learning Semantically Consistent Deformable 3D Representations from In-the-Wild Imagescited by 0
  9. 2026Category-Level 3D Correspondence in Camera Space via Morphable Object Priorscited by 0
  10. 2026Every9D-21M: Large-Scale Real-World 9D Canonicalization of Everyday Objectscited by 0
  11. 2026SOCO: Benchmarking Semantic Object Correspondence in Vision Foundation Modelscited by 0
  12. 2026Emergence of a Shared Canonical Object Frame from In-the-Wild Videoscited by 0
1 more
  1. 2026ImCorr: Sub-pixel Semantic Correspondence via Implicit Feature Decodingcited by 0