mlsanityData to 14 Oct 2026

Sapiens2

arXiv 2604.21681 ↗First posted 23 Apr 2026cs.CVcited by 4

Not on the map. Why?

Abstract (arXiv)

We present Sapiens2, a model family of high-resolution transformers for human-centric vision focused on generalization, versatility, and high-fidelity outputs. Our model sizes range from 0.4 to 5 billion parameters, with native 1K resolution and hierarchical variants that support 4K. Sapiens2 substantially improves over its predecessor in both pretraining and post-training. First, to learn features that capture low-level details (for dense prediction) and high-level semantics (for zero-shot or few-label settings), we combine masked image reconstruction with self-distilled contrastive objectives. Our evaluations show that this unified pretraining objective is better suited for a wider range of downstream tasks. Second, along the data axis, we pretrain on a curated dataset of 1 billion high-quality human images and improve the quality and quantity of task annotations. Third, architecturally, we incorporate advances from frontier models that enable longer training schedules with improved stability. Our 4K models adopt windowed attention to reason over longer spatial context and are pretrained with 2K output resolution. Sapiens2 sets a new state-of-the-art and improves over the first generation on pose (+4 mAP), body-part segmentation (+24.3 mIoU), normal estimation (45.6% lower angular error) and extends to new tasks such as pointmap and albedo estimation. Code: https://github.com/facebookresearch/sapiens2

Key facts. Cited by 4 papers in the map (data through 14 Oct 2026). Builds on 56 of its 63 references that are in the map. First posted on arXiv 23 Apr 2026.

Builds on

56 of 63 references.

References · 56

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  3. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 14,132
  4. 2020A Simple Framework for Contrastive Learning of Visual Representationscited by 12,049
  5. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,498
  6. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,974
  7. 2021Masked Autoencoders Are Scalable Vision Learnerscited by 7,403
  8. 2021Emerging Properties in Self-Supervised Vision Transformerscited by 6,968
  9. 2020Scaling Laws for Neural Language Modelscited by 6,586
  10. 2020Bootstrap your own latent: A new approach to self-supervised Learningcited by 4,750
  11. 2021BEiT: BERT Pre-Training of Image Transformerscited by 2,375
  12. 2020GLU Variants Improve Transformercited by 1,804
38 more
  1. 2021An Empirical Study of Training Self-Supervised Vision Transformerscited by 1,616
  2. 2023DUSt3R: Geometric 3D Vision Made Easycited by 1,567
  3. 2025VGGT: Visual Geometry Grounded Transformercited by 1,539
  4. 2019Root Mean Square Layer Normalizationcited by 1,480
  5. 2025DINOv3cited by 1,414
  6. 2020Big Self-Supervised Models are Strong Semi-Supervised Learnerscited by 1,380
  7. 2023GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpointscited by 1,346
  8. 2016Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Networkcited by 1,204
  9. 2023Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecturecited by 1,039
  10. 2021iBOT: Image BERT Pre-Training with Online Tokenizercited by 903
  11. 2025V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planningcited by 809
  12. 2023Scaling Vision Transformers to 22 Billion Parameterscited by 694
  13. 2024Revisiting Feature Prediction for Learning Visual Representations from Videocited by 531
  14. 2021MViTv2: Improved Multiscale Vision Transformers for Classification and Detectioncited by 519
  15. 2024Depth Pro: Sharp Monocular Metric Depth in Less Than a Secondcited by 404
  16. 2024UniDepth: Universal Monocular Metric Depth Estimationcited by 382
  17. 2023Metric3D: Towards Zero-shot Metric 3D Prediction from A Single Imagecited by 365
  18. 2017Cascaded Pyramid Network for Multi-Person Pose Estimationcited by 341
  19. 2025Perception Encoder: The best visual embeddings are not at the output of the networkcited by 326
  20. 2024MoGe: Unlocking Accurate Monocular Geometry Estimation for Open-Domain Images with Optimal Training Supervisioncited by 299
  21. 2023Hiera: A Hierarchical Vision Transformer without the Bells-and-Whistlescited by 290
  22. 2020Query-Key Normalization for Transformerscited by 280
  23. 2017ICNet for Real-Time Semantic Segmentation on High-Resolution Imagescited by 261
  24. 2024Sapiens: Foundation for Human Vision Modelscited by 229
  25. 2020Deep Learning-Based Human Pose Estimation: A Surveycited by 207
  26. 2022A Comprehensive Survey of Few-shot Learning: Evolution, Applications, Challenges, and Opportunitiescited by 163
  27. 2022Contrastive Masked Autoencoders are Stronger Vision Learnerscited by 133
  28. 2024Rethinking Inductive Biases for Surface Normal Estimationcited by 124
  29. 2022ConvMAE: Masked Convolution Meets Masked Autoencoderscited by 105
  30. 2023Beyond Appearance: a Semantic Controllable Self-Supervised Learning Framework for Human-Centric Visual Taskscited by 96
  31. 2022How Mask Matters: Towards Theoretical Understandings of Masked Autoencoderscited by 86
  32. 2023Semantic Segmentation using Vision Transformers: A surveycited by 73
  33. 2024SwitchLight: Co-design of Physics-driven Architecture and Pre-training Framework for Human Portrait Relightingcited by 67
  34. 2025Meta CLIP 2: A Worldwide Scaling Recipecited by 64
  35. 2017Checkerboard artifact free sub-pixel convolution: A note on sub-pixel convolution, resize convolution and convolution resizecited by 37
  36. 2023Loss Functions in the Era of Semantic Segmentation: A Survey and Outlookcited by 34
  37. 2023HAP: Structure-Aware Masked Image Modeling for Human-Centric Perceptioncited by 33
  38. 2024Masked Image Modeling: A Surveycited by 29

50 most cited of 56 All 56 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 4.

4 papers.

  1. 2026Biomechanical 3D Body: Self-Supervised Distillation of Biomechanical Pose from a 3D Body Foundation Modelcited by 1
  2. 2026Cross-Model Distillation of a Human-Pose Foundation Model from Unannotated Infant Video for Markerless 3D Pose Estimationcited by 0
  3. 2026SynthGait-19K: A Physically Grounded Synthetic Video Dataset for Gait Parameter Estimationcited by 0
  4. 2026PACT: End-to-End Learning of Human Pose, Contacts, and Forces from Videocited by 0