Pith. sign in

REVIEW 3 major objections 4 minor 66 references

SAT-HMR: Real-Time Multi-Person 3D Mesh Estimation via Scale-Adaptive Tokens

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read Scale-adaptive tokens—high-res for small people, low-res for large, pooled background—let a DETR-style mesh estimator keep high-resolution accuracy at 24 FPS.

desk verdict A genuine efficiency win with a slightly over-stated 'preserves accuracy' claim; the scale-adaptive token mechanism is new, well-ablated, and worth refereeing. read the letter →

arxiv 2411.19824 v4 pith:3GALTEC5 submitted 2024-11-29 cs.CV

classification cs.CV
keywords 3Dhumanmeshestimationmulti-personscale-adaptivetokensDETRreal-timeinferenceSMPLtokenpruningsingleRGBimage
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper is trying to establish that the accuracy benefit of high-resolution input in one-stage multi-person 3D mesh estimation can be preserved without processing the entire image at high resolution. Its proposal, scale-adaptive tokens, predicts a patch-level scale map and spends resolution where it matters: small-scale people (far from the camera, children, or in closed poses) get high-resolution tokens, large-scale people keep low-resolution tokens, and background regions are pooled into fewer tokens. On the AGORA test set the method reports 63.3 mm mean vertex error at 42 ms per image, comparable to Multi-HMR's 61.1 mm at 231.7 ms with 1288-resolution input, and it runs at roughly 24 FPS. If the claim holds, real-time multi-person mesh recovery no longer has to sacrifice accuracy on small or distant people to stay fast.

What carries the argument

The load-bearing object is the scale-adaptive token set $T_{\text{SA}} = \{T'_B, T_{\text{LARGE}}, T_{\text{HR}}\}$, assembled from a predicted patch-level scale map $S(i,j) = (c,s)$. Here $c$ is the confidence that a patch overlaps a person and $s$ is the diagonal of that person's bounding box normalized by the longer side of the high-resolution image. A shallow three-layer transformer and an MLP scale head predict the map from low-resolution tokens; patches below the scale threshold $\alpha_s = 0.5$ are re-tokenized at four times the resolution by a parallel shallow encoder, large-scale patches remain low-resolution, and background tokens are pooled four-to-one to cut their number. This conditional allocation is what carries the argument: it concentrates computation on the difficult small-scale cases that previously forced one-stage methods to run the whole image at high resolution.

What would settle it

Take the trained model, swap in the ground-truth scale map at test time, and compare per-scale-range mean vertex error on AGORA validation; if mispredicted small-scale patches show errors near the low-resolution baseline, the scale predictor is the bottleneck, whereas if the gap stays small (the BEDLAM ablation shows only 0.2 mm average), the residual error lies in mesh regression rather than token allocation.

Watch

Extended reading notes

Core claim

The central discovery is that uniform high-resolution tokenization of the whole image is wasteful because different people need different levels of detail. The paper's own baseline analysis shows that raising input resolution from 518 to 1288 cuts mean vertex error by nearly 35 mm for individuals in the 0-10% scale range, while giving almost nothing to people covering more than 30% of the image. SAT-HMR therefore builds a scale-adaptive token set from a patch-level scale map: patches classified as small-scale are replaced by high-resolution tokens extracted from a 1288-resolution image, large-scale patches keep base 644-resolution tokens, and background tokens are spatially pooled. These tokens feed a transformer encoder and a DAB-DETR-style decoder that regresses SMPL pose, shape, and translation. The result is accuracy close to high-resolution state-of-the-art models at a fraction of the cost—133.1 G MACs and 42 ms per image versus 6104.6 G and 231.7 ms for Multi-HMR at 1288 resolution.

Load-bearing premise

The load-bearing premise is that the learned patch-level scale map correctly identifies which image patches contain small-scale people, so high-resolution tokens are placed where they matter; if the scale prediction is wrong for a person, that person effectively gets only low-resolution features and the accuracy gain disappears.

Editorial extensions

If this is right

  • One-stage mesh estimators can run at real-time speed without dropping to low-resolution accuracy: 63.3 mm MVE on AGORA test at 42 ms per image, within 2.2 mm of Multi-HMR's 1288-resolution result while running about 5.5 times faster.
  • Because the change is confined to token construction in the encoder, the scale-adaptive recipe can be transplanted into other DETR-style pose and mesh estimators without altering their decoders or losses.
  • Background tokens carry context that matters: pooling them preserves accuracy for large or truncated people, while dropping them entirely hurts the largest-scale range on BEDLAM.
  • Scale prediction is accurate enough in aggregate (0.98 F1 on BEDLAM) that replacing it with ground-truth scale changes average MVE only from 56.0 to 55.8 mm, so most remaining error is in mesh regression rather than token allocation.
  • Increasing the base resolution of the adaptive scheme from 644* to 896* continues to cut small-scale error (from 84.6 to 76.5 mm for 0-10% scale in AGORA validation), so the approach gains from extra compute without changing the mechanism.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because the model is not age- or height-aware, children and atypically small adults can be mis-scaled; adding an explicit height or age prior to the scale head or depth head is a natural testable extension that targets the failure mode the paper itself illustrates.
  • The 35 mm gap between low- and high-resolution baselines for small-scale individuals suggests small-instance feature detail, not depth ambiguity, is the dominant bottleneck in one-stage mesh estimation; the same adaptive allocation could be applied to face and hand regions for whole-body estimation, which the authors name as future work.
  • The scale map is predicted from only three transformer layers on low-resolution tokens, so its 0.98 F1 may not transfer to unusual camera angles or extremely dense crowds; stress-testing on those distributions would show whether allocation or regression deserves the next round of capacity.
  • Since the ground-truth scale map closes only about 0.2 mm of average error, spending the saved compute on stronger decoders or SMPL priors may buy more accuracy per FLOP than further improving scale prediction.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes SAT-HMR, a one-stage DETR-style multi-person 3D human mesh estimator that replaces uniformly partitioned image tokens with scale-adaptive tokens. A patch-level scale map predicted from low-resolution tokens classifies patches into background, small-scale, and large-scale categories; small-scale patches are replaced by higher-resolution tokens, background tokens are pooled, and large-scale tokens are kept unchanged. Experiments on AGORA, 3DPW, MuPoTS, and CMU Panoptic report performance comparable to recent SOTA methods at about 24 FPS (42 ms) with 133 G MACs, versus 4-5 FPS for high-resolution baselines. The paper includes ablations on BEDLAM and AGORA validation analyzing per-scale accuracy, background pooling, the scale threshold, and the accuracy of the scale map.

Significance. If the speed-accuracy tradeoff is taken as the contribution, the paper is a solid engineering contribution: it demonstrates a controlled speedup from about 5 FPS to 24 FPS at a modest accuracy cost relative to the full-resolution baseline (e.g., 53.2 vs 56.0 MVE on BEDLAM), with a thorough per-scale analysis and a public code release. The central conceptual claim, however, that the method 'preserves the accuracy benefits of high-resolution processing' is not supported by the authors' own comparisons; the method incurs a systematic MVE gap of roughly 3-4 mm relative to the full-resolution baseline. The paper is strongest when framed as an efficiency-accuracy tradeoff and weakest when framed as accuracy-preserving. The empirical design is not circular: the scale threshold is a hyperparameter tuned on validation data, and the GT-scale-map ablation (Table B5) shows that scale prediction is accurate (F1=0.98) and not a bottleneck.

major comments (3)
  1. [§4.4, Tables A1 and B2] The claim in the Abstract and in the Contributions that the method 'preserves the accuracy benefits of high-resolution processing' is not supported by the authors' own single-resolution baselines. On BEDLAM validation, the Res-1288 baseline achieves 53.2 MVE versus 56.0 for the proposed 644* model (Table A1); on AGORA validation, the corresponding numbers are 61.9 versus 65.5 (Table B2). These gaps are comparable to the differences among SOTA methods in Table 1 (e.g., 61.1 for Multi-HMR 1288 versus 63.3 for Ours), so the accuracy cost is not negligible. The main AGORA test comparison in Table 1 does not include the authors' own ViT-B Res-1288 baseline, so the headline 'comparable to SOTA' conflates the token-allocation mechanism with backbone and output-representation choices. The paper should either report the own-baseline result on the AGORA test set or explicitly recast the claim as a speed-accuracy tradeoff with a modest, scale-dependent accuracy cost.
  2. [Table B2] The per-scale breakdown in Table B2 shows that the largest absolute gap between the proposed 644* model and the Res-1288 baseline occurs for large-scale individuals, not small-scale ones. On AGORA validation, the 30%+ scale range shows MVE of 52.7 for Ours versus 48.2 for Res-1288, a 4.5 mm gap, while the 0-10% range shows 84.6 versus 82.2, a 2.4 mm gap. For large individuals, Ours essentially matches the low-resolution Res-644 baseline (53.0 MVE) rather than preserving the high-resolution benefit. This weakens the motivation stated in §1 that high-resolution input gives only marginal improvements for large-scale individuals, and it raises a question about where the resolution benefit is actually lost. Please discuss this per-scale behavior explicitly and, if possible, provide an analysis of why the large-scale range loses the most accuracy.
  3. [§4.3, Table 1] The efficiency comparison to Multi-HMR is not apples-to-apples: Multi-HMR uses a ViT-L backbone and regresses SMPL-X, while the proposed method uses ViT-B and regresses SMPL. The 5x speedup claim in the Contributions therefore conflates the scale-adaptive token mechanism with architecture and output-representation choices. The controlled comparison in Fig. 2 and Table A1 (same architecture, different resolutions) is the appropriate evidence for the efficiency-accuracy tradeoff, and the SOTA table should be interpreted with this caveat. The paper should state this limitation directly when reporting the speedup.
minor comments (4)
  1. [Sec. B.3] The paper's own limitation statement that the method is not age- or height-aware is directly relevant to the small-scale population the method is designed to handle; Fig. B5 shows a child depth error. This limitation should be moved to the main text, since it tempers the claim about small-scale individuals.
  2. [Supplementary, Table B4 paragraph] The word 'interence' appears in the sentence 'decreases efficiency with longer interence time cost'; it should be 'inference'.
  3. [Sec. 3.4 and Sec. A.2] There is a missing space in 'set toλmap = 4' in Sec. A.2; please fix the formatting of the loss weights.
  4. [§4.3, last paragraph] The phrase 'achieving an unparalleled balance' is promotional and not supported by the data, since the method sacrifices a few mm of accuracy relative to the full-resolution baseline; please replace it with a neutral statement.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the scale-adaptive token mechanism is an architectural allocation scheme, and the accuracy/efficiency claims are empirical and tested against held-out benchmarks.

full rationale

The paper's central mechanism is scale-adaptive token selection: a patch-level scale map defined by Eq. (1) from ground-truth bounding-box diagonals is predicted from low-resolution tokens, and this map decides which low-resolution tokens are replaced by high-resolution patches. This is an internal routing/allocation mechanism, not a derivation whose output is defined in terms of its own inputs. The claimed preservation of high-resolution accuracy is evaluated empirically on held-out AGORA test, 3DPW test, MuPoTS, and CMU Panoptic; the headline 'preserves the accuracy benefits of high-resolution processing' is supported by comparisons against single-resolution baselines in ablations (Table A1 and Table B2), and the scale-map predictor is independently checked by replacing it with ground truth (Table B5), which changes average MVE only slightly (56.0 to 55.8 on BEDLAM). The scale threshold αs is tuned on BEDLAM validation (Table B4), but hyperparameter selection on a validation split is standard practice, not a fitted input renamed as a prediction. The paper's self-citations (e.g., [26, 27, 50, 59]) appear only as related-work context and are not load-bearing for the proposed method, and no uniqueness theorem or prior result by the same authors is invoked to force the design. The acknowledged limitations (lack of height/age awareness, body-only estimation, larger depth errors for children) and the absence of the authors' own full-resolution ViT-B baseline on the AGORA test set are benchmarking/completeness concerns, not circularity. The derivation chain is therefore self-contained with respect to its inputs: the proposed method is an empirical engineering contribution whose claims are tested against external benchmarks and internal ablations, so no circular step is present.

Assumptions & free parameters 4 free parameters · 5 assumptions · 2 invented entities

The paper introduces no new physical entities. The scale-adaptive mechanism relies on a learned scale map that is validated against GT, good. The main free parameters are the thresholds and resolution choices, tuned on validation sets. The key axioms are standard domain assumptions of the field (SMPL, pinhole camera, pseudo-labels).

free parameters (4)
  • Scale threshold αs = 0.5
    Chosen from BEDLAM validation sweep (Table B4); smaller values reduce high-res coverage and larger values slow inference or destabilize training. This is a hand-tuned trade-off, not derived.
  • Background confidence threshold αc = 0.3
    Determines which low-res tokens are classified as background and pooled. No ablation reported for this threshold.
  • Input base / high-res resolutions = 644 base, 1288 max
    Sets the token budget; Table B2 shows 644* is a chosen speed-accuracy balance, with 896* more accurate but slower.
  • Loss weights (λmap, λdepth, λpose, λshape, λj3d, λj2d, λbox, λdet) = 4, 0.5, 5, 3, 8, 40, 2, 4
    Follows prior DETR-style HMR losses (Sec A.2); not the central claim.
assumptions (5)
  • domain assumption SMPL represents the human body sufficiently for the target application
    The paper regresses SMPL pose/shape parameters (Sec 3.1) and treats mesh vertices as the output space; this is a domain assumption shared with the field.
  • domain assumption Pinhole camera model with fixed 60 degree FOV
    Used for projecting 3D joints to 2D and for the depth loss (Sec A.1); real images with different FOV will introduce bias. Assumed to be acceptable for the benchmarks.
  • domain assumption NeuralAnnot pseudo-GT is reliable for real-world 2D supervision
    COCO, CrowdPose, and MPII are trained with pseudo labels from NeuralAnnot (Sec 4.1); errors in these labels are assumed not to substantially degrade the model.
  • domain assumption DINOv2 pretrained features transfer to mesh regression
    Encoder initialized with DINOv2 (Sec 4.2); the method assumes these features, along with added layers, are a suitable representation for the scale map and mesh decoding.
  • ad hoc to paper GT-derived scale map is a valid training target
    The scale map is supervised with a scale computed from GT bounding boxes (Eq. 1), defining 'resolution needed' in terms of bbox diagonal. This is a heuristic specific to the paper's design.
invented entities (2)
  • Patch-level scale map S independent evidence
    purpose: Per-patch prediction of person confidence and relative scale to guide token resolution assignment (Sec 3.3)
    Evaluated directly against GT-derived scale maps via F1 (0.98) and MAE (0.056) on BEDLAM val (Table B5); this is a falsifiable internal representation, though not a physical entity.
  • Scale-adaptive tokens TSA
    purpose: Mixed-resolution token sequence that replaces uniform tokens in the encoder (Sec 3.3)
    A model component without an external falsifiable handle; its benefit is measured only through downstream MVE.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SAT-HMR: Real-Time Multi-Person 3D Mesh Estimation via Scale-Adaptive Tokens." pith.science (2026). https://pith.science/paper/3GALTEC5

@misc{pith2026241119824,
  author       = {Pith},
  title        = {Pith review of: SAT-HMR: Real-Time Multi-Person 3D Mesh Estimation via Scale-Adaptive Tokens},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3GALTEC5}},
  note         = {Machine review of arXiv:2411.19824}
}
read the original abstract

We propose a one-stage framework for real-time multi-person 3D human mesh estimation from a single RGB image. While current one-stage methods, which follow a DETR-style pipeline, achieve state-of-the-art (SOTA) performance with high-resolution inputs, we observe that this particularly benefits the estimation of individuals in smaller scales of the image (e.g., those far from the camera), but at the cost of significantly increased computation overhead. To address this, we introduce scale-adaptive tokens that are dynamically adjusted based on the relative scale of each individual in the image within the DETR framework. Specifically, individuals in smaller scales are processed at higher resolutions, larger ones at lower resolutions, and background regions are further distilled. These scale-adaptive tokens more efficiently encode the image features, facilitating subsequent decoding to regress the human mesh, while allowing the model to allocate computational resources more effectively and focus on more challenging cases. Experiments show that our method preserves the accuracy benefits of high-resolution processing while substantially reducing computational cost, achieving real-time inference with performance comparable to SOTA methods.

Figures

Figures reproduced from arXiv: 2411.19824 by the authors.

Figure 1
Figure 1. (a) We propose scale-adaptive tokens in our one-stage framework for real-time multi-person 3D mesh estimation. Our method introduces scale-adaptive tokens, dynamically adjusted based on the relative size of individuals in the image, to more efficiently encode features, enabling real-time and accurate multi-person mesh estimation. We present a conceptual visualization of the scale-adaptive tokens. The right column vi… view at source ↗
Figure 2
Figure 2. Estimation errors and FPS of baselines with different resolutions and our method across individuals at various scales. The scale of an individual refers to the person’s size relative to the overall image and please refer to Sec. 3.3 for mathematical definition. The colored lines show the Mean Vertex Error (MVE) errors (left y-axis) of the baselines with different resolutions (Res.) on the AGORA [33] validation set. … view at source ↗
Figure 3
Figure 3. Overview of (top) the baseline method and (bottom) our method with scale-adaptive tokens. Top: Our baseline method adopts a DETR-style [5] pipeline consisting of a Transformer encoder, decoder, and prediction heads for regressing SMPL parameters. Bottom: Our method focuses on efficient feature encoding using scale-adaptive tokens. Specifically, low-resolution and high-resolution patches are extracted from the input … view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Comparison with SOTA methods [2, 44–46] on in-the-wild images from the Internet. Red dashed circles highlight areas with incorrect estimations. The third case is left blank due to the small scale of individuals. Please zoom in for details [PITH_FULL_IMAGE:figures/full…
Figure 5
Figure 5. Figure 5: Qualitative comparison of different resolutions for our baseline. Resolution for the baseline: (a) 518, (b) 1288. Red dashed circles highlight differences; zoom in for details. tion performance of our method on the MuPoTS and CMU Panoptic datasets, as shown in Tab. 2 a…
Figure 6
Figure 6. Figure 6: Visualization of our predicted patch-level scale maps S and scale-adaptive tokens TSA. We show the predicted scale maps, scale-adaptive tokens, and the estimated meshes overlaid on the image. In the scale map, colors represent scale values, with uncolored areas indicat…
Figure 7
Figure 7. Figure 7: Qualitative results of our method. We visualize four cases from AGORA [33], BEDLAM [3], CMU Panoptic [15] and 3DPW [49] datasets, displaying the input, the estimated mesh overlay, and an elevated view (top to bottom). the significance of background context for estimati…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

66 extracted references · 58 canonical work pages

  1. [1]

    2d human pose estimation: New benchmark and state of the art analysis

    Mykhaylo Andriluka, Leonid Pishchulin, Peter Gehler, and Bernt Schiele. 2d human pose estimation: New benchmark and state of the art analysis. In CVPR, pages 3686–3693,

  2. [2]

    Multi-hmr: Multi-person whole-body human mesh recovery in a single shot

    Fabien Baradel, Matthieu Armando, Salma Galaaoui, Romain Brégier, Philippe Weinzaepfel, Grégory Rogez, and Thomas Lucas. Multi-hmr: Multi-person whole-body human mesh recovery in a single shot. In ECCV, 2024. 2, 3, 4, 6, 7, 1, 5

  3. [3]

    Bedlam: A synthetic dataset of bodies exhibiting de- tailed lifelike animated motion

    Michael J Black, Priyanka Patel, Joachim Tesch, and Jinlong Yang. Bedlam: A synthetic dataset of bodies exhibiting de- tailed lifelike animated motion. In CVPR, pages 8726–8737,

  4. [4]

    Token merging: Your ViT but faster

    Daniel Bolya, Cheng-Yang Fu, Xiaoliang Dai, Peizhao Zhang, Christoph Feichtenhofer, and Judy Hoffman. Token merging: Your ViT but faster. InICLR, 2023. 3

  5. [5]

    End- to-end object detection with transformers

    Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, and Sergey Zagoruyko. End- to-end object detection with transformers. In ECCV, pages 213–229. Springer, 2020. 2, 3, 4, 5, 1

  6. [6]

    Cross- attention of disentangled modalities for 3d human mesh re- covery with transformers

    Junhyeong Cho, Kim Youwang, and Tae-Hyun Oh. Cross- attention of disentangled modalities for 3d human mesh re- covery with transformers. In ECCV, 2022. 3

  7. [7]

    Learning to estimate robust 3d human mesh from in-the-wild crowded scenes

    Hongsuk Choi, Gyeongsik Moon, JoonKyu Park, and Ky- oung Mu Lee. Learning to estimate robust 3d human mesh from in-the-wild crowded scenes. InCVPR, pages 1475–1484,

  8. [8]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. In ICLR, 2021. 3

Show all 66 references
  1. [9]

    Tore: Token reduction for efficient human mesh recov- ery with transformer

    Zhiyang Dou, Qingxuan Wu, Cheng Lin, Zeyu Cao, Qiangqiang Wu, Weilin Wan, Taku Komura, and Wenping Wang. Tore: Token reduction for efficient human mesh recov- ery with transformer. In ICCV, pages 15143–15155, 2023. 3

  2. [10]

    Cam-convs: Camera-aware multi-scale convolutions for single-view depth

    Jose M Facil, Benjamin Ummenhofer, Huizhong Zhou, Luis Montesano, Thomas Brox, and Javier Civera. Cam-convs: Camera-aware multi-scale convolutions for single-view depth. In CVPR, pages 11826–11835, 2019. 5

  3. [11]

    Humans in 4d: Recon- structing and tracking humans with transformers

    Shubham Goel, Georgios Pavlakos, Jathushan Rajasegaran, Angjoo Kanazawa, and Jitendra Malik. Humans in 4d: Recon- structing and tracking humans with transformers. In ICCV, pages 14783–14794, 2023. 3, 1

  4. [12]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. InCVPR, pages 770–778, 2016. 3

  5. [13]

    Catalin Ionescu, Dragos Papava, Vlad Olaru, and Cristian Sminchisescu. Human3. 6m: Large scale datasets and predic- tive methods for 3d human sensing in natural environments. IEEE TPAMI, 36(7):1325–1339, 2013. 6, 1

  6. [14]

    Coherent reconstruction of multiple humans from a single image

    Wen Jiang, Nikos Kolotouros, Georgios Pavlakos, Xiaowei Zhou, and Kostas Daniilidis. Coherent reconstruction of multiple humans from a single image. In CVPR, pages 5579– 5588, 2020. 2, 3, 6, 7

  7. [15]

    Panoptic studio: A massively multiview system for social motion capture

    Hanbyul Joo, Hao Liu, Lei Tan, Lin Gui, Bart Nabbe, Iain Matthews, Takeo Kanade, Shohei Nobuhara, and Yaser Sheikh. Panoptic studio: A massively multiview system for social motion capture. In ICCV, pages 3334–3342, 2015. 6, 7, 8, 2, 3

  8. [16]

    Learning to reconstruct 3d human pose and shape via model-fitting in the loop

    Nikos Kolotouros, Georgios Pavlakos, Michael J Black, and Kostas Daniilidis. Learning to reconstruct 3d human pose and shape via model-fitting in the loop. In ICCV, pages 2252–2261, 2019. 1

  9. [17]

    Crowdpose: Efficient crowded scenes pose estimation and a new benchmark

    Jiefeng Li, Can Wang, Hao Zhu, Yihuan Mao, Hao-Shu Fang, and Cewu Lu. Crowdpose: Efficient crowded scenes pose estimation and a new benchmark. In CVPR, pages 10863– 10872, 2019. 6, 1

  10. [18]

    End-to-end human pose and mesh reconstruction with transformers

    Kevin Lin, Lijuan Wang, and Zicheng Liu. End-to-end human pose and mesh reconstruction with transformers. In CVPR, pages 1954–1963, 2021. 3

  11. [19]

    Mesh graphormer

    Kevin Lin, Lijuan Wang, and Zicheng Liu. Mesh graphormer. In ICCV, pages 12939–12948, 2021. 3

  12. [20]

    Microsoft coco: Common objects in context

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Dollár, and C Lawrence Zitnick. Microsoft coco: Common objects in context. In ECCV, pages 740–755, 2014. 6, 1

  13. [21]

    Focal loss for dense object detection

    Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Dollar. Focal loss for dense object detection. In ICCV,

  14. [22]

    Group pose: A simple baseline for end-to-end multi-person pose estimation

    Huan Liu, Qiang Chen, Zichang Tan, Jiang-Jiang Liu, Jian Wang, Xiangbo Su, Xiaolong Li, Kun Yao, Junyu Han, Errui Ding, et al. Group pose: A simple baseline for end-to-end multi-person pose estimation. In ICCV, pages 15029–15038,

  15. [23]

    DAB-DETR: Dynamic anchor boxes are better queries for DETR

    Shilong Liu, Feng Li, Hao Zhang, Xiao Yang, Xianbiao Qi, Hang Su, Jun Zhu, and Lei Zhang. DAB-DETR: Dynamic anchor boxes are better queries for DETR. In ICLR, 2022. 3, 5, 1, 2

  16. [24]

    Smpl: A skinned multi- person linear model

    Matthew Loper, Naureen Mahmood, Javier Romero, Gerard Pons-Moll, and Michael J Black. Smpl: A skinned multi- person linear model. ACM TOG, 34(6):1–16, 2015. 2, 3, 6, 1

  17. [25]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. In ICLR, 2017. 1

  18. [26]

    3d human mesh estimation from virtual mark- ers

    Xiaoxuan Ma, Jiajun Su, Chunyu Wang, Wentao Zhu, and Yizhou Wang. 3d human mesh estimation from virtual mark- ers. In CVPR, pages 534–543, 2023. 3

  19. [27]

    Vmarker-pro: Probabilistic 3d human mesh estimation from virtual markers

    Xiaoxuan Ma, Jiajun Su, Yuan Xu, Wentao Zhu, Chunyu Wang, and Yizhou Wang. Vmarker-pro: Probabilistic 3d human mesh estimation from virtual markers. IEEE Trans- actions on Pattern Analysis and Machine Intelligence, 2025. 3

  20. [28]

    Monocular 3d human pose estimation in the wild using im- proved cnn supervision

    Dushyant Mehta, Helge Rhodin, Dan Casas, Pascal Fua, Olek- sandr Sotnychenko, Weipeng Xu, and Christian Theobalt. Monocular 3d human pose estimation in the wild using im- proved cnn supervision. In 3DV, pages 506–516, 2017. 6 9

  21. [29]

    Single-shot multi-person 3d pose estimation from monocular rgb

    Dushyant Mehta, Oleksandr Sotnychenko, Franziska Mueller, Weipeng Xu, Srinath Sridhar, Gerard Pons-Moll, and Chris- tian Theobalt. Single-shot multi-person 3d pose estimation from monocular rgb. In 3DV, pages 120–130, 2018. 6, 1, 3

  22. [30]

    Adavit: Adaptive vision transformers for efficient image recognition

    Lingchen Meng, Hengduo Li, Bor-Chun Chen, Shiyi Lan, Zuxuan Wu, Yu-Gang Jiang, and Ser-Nam Lim. Adavit: Adaptive vision transformers for efficient image recognition. In CVPR, pages 12309–12318, 2022. 3

  23. [31]

    Neu- ralannot: Neural annotator for 3d human mesh training sets

    Gyeongsik Moon, Hongsuk Choi, and Kyoung Mu Lee. Neu- ralannot: Neural annotator for 3d human mesh training sets. In CVPR, pages 2299–2307, 2022. 6, 1

  24. [32]

    Dinov2: Learning robust visual features without supervision

    Maxime Oquab, Timothée Darcet, Théo Moutakanni, Huy V V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel HAZIZA, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without supervision. Transactions on Machine Learning Research, 2024. 6

  25. [33]

    Agora: Avatars in geography optimized for regression analysis

    Priyanka Patel, Chun-Hao P Huang, Joachim Tesch, David T Hoffmann, Shashank Tripathi, and Michael J Black. Agora: Avatars in geography optimized for regression analysis. In CVPR, pages 13468–13478, 2021. 1, 2, 6, 7, 8, 3, 4, 5

  26. [34]

    Expressive body capture: 3d hands, face, and body from a single image

    Georgios Pavlakos, Vasileios Choutas, Nima Ghorbani, Timo Bolkart, Ahmed AA Osman, Dimitrios Tzionas, and Michael J Black. Expressive body capture: 3d hands, face, and body from a single image. In CVPR, pages 10975–10985,

  27. [35]

    Dynamic graph reasoning for multi-person 3d pose estimation

    Zhongwei Qiu, Qiansheng Yang, Jian Wang, and Dongmei Fu. Dynamic graph reasoning for multi-person 3d pose estimation. In Proceedings of the 30th ACM International Conference on Multimedia, pages 3521–3529, 2022. 2, 3

  28. [36]

    Psvt: End-to-end multi-person 3d pose and shape estimation with progressive video transformers

    Zhongwei Qiu, Qiansheng Yang, Jian Wang, Haocheng Feng, Junyu Han, Errui Ding, Chang Xu, Dongmei Fu, and Jingdong Wang. Psvt: End-to-end multi-person 3d pose and shape estimation with progressive video transformers. In CVPR, pages 21254–21263, 2023. 3, 6, 7

  29. [37]

    Dynamicvit: Efficient vision trans- formers with dynamic token sparsification

    Yongming Rao, Wenliang Zhao, Benlin Liu, Jiwen Lu, Jie Zhou, and Cho-Jui Hsieh. Dynamicvit: Efficient vision trans- formers with dynamic token sparsification. In NeurIPS, pages 13937–13949, 2021. 3

  30. [38]

    Faster R-CNN: Towards real-time object detection with region proposal networks

    Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster R-CNN: Towards real-time object detection with region proposal networks. In NeurIPS, 2015. 3

  31. [39]

    Learn- ing to merge tokens in vision transformers

    Cedric Renggli, André Susano Pinto, Neil Houlsby, Basil Mustafa, Joan Puigcerver, and Carlos Riquelme. Learn- ing to merge tokens in vision transformers. arXiv preprint arXiv:2202.12015, 2022. 3

  32. [40]

    Generalized in- tersection over union: A metric and a loss for bounding box regression

    Hamid Rezatofighi, Nathan Tsoi, JunYoung Gwak, Amir Sadeghian, Ian Reid, and Silvio Savarese. Generalized in- tersection over union: A metric and a loss for bounding box regression. In CVPR, pages 658–666, 2019. 5

  33. [41]

    Vision trans- formers with mixed-resolution tokenization

    Tomer Ronen, Omer Levy, and Avram Golbert. Vision trans- formers with mixed-resolution tokenization. In CVPRW, pages 4613–4622, 2023. 3

  34. [42]

    Robots that can see: Leveraging human pose for trajectory prediction

    Tim Salzmann, Hao-Tien Lewis Chiang, Markus Ryll, Dorsa Sadigh, Carolina Parada, and Alex Bewley. Robots that can see: Leveraging human pose for trajectory prediction. IEEE Robotics and Automation Letters, 2023. 2

  35. [43]

    End-to-end multi-person pose estimation with transformers

    Dahu Shi, Xing Wei, Liangqi Li, Ye Ren, and Wenming Tan. End-to-end multi-person pose estimation with transformers. In CVPR, pages 11069–11078, 2022. 3

  36. [44]

    Aios: All-in-one-stage expressive human pose and shape estimation

    Qingping Sun, Yanjun Wang, Ailing Zeng, Wanqi Yin, Chen Wei, Wenjia Wang, Haiyi Mei, Chi-Sing Leung, Ziwei Liu, Lei Yang, et al. Aios: All-in-one-stage expressive human pose and shape estimation. In CVPR, pages 1834–1843, 2024. 2, 3, 4, 6, 7, 1, 5

  37. [45]

    Monocular, one-stage, regression of multiple 3d people

    Yu Sun, Qian Bao, Wu Liu, Yili Fu, Michael J Black, and Tao Mei. Monocular, one-stage, regression of multiple 3d people. In ICCV, pages 11179–11188, 2021. 3, 6, 7, 1, 2

  38. [46]

    Putting people in their place: Monocular regression of 3d people in depth

    Yu Sun, Wu Liu, Qian Bao, Yili Fu, Tao Mei, and Michael J Black. Putting people in their place: Monocular regression of 3d people in depth. In CVPR, pages 13243–13252, 2022. 2, 3, 6, 7, 1, 4

  39. [47]

    Quadtree attention for vision transformers

    Shitao Tang, Jiahui Zhang, Siyu Zhu, and Ping Tan. Quadtree attention for vision transformers. In ICLR, 2022. 3

  40. [48]

    Attention is all you need

    A Vaswani. Attention is all you need. NeurIPS, 2017. 3

  41. [49]

    Recovering accurate 3d human pose in the wild using imus and a moving camera

    Timo von Marcard, Roberto Henschel, Michael J Black, Bodo Rosenhahn, and Gerard Pons-Moll. Recovering accurate 3d human pose in the wild using imus and a moving camera. In ECCV, pages 601–617, 2018. 1, 6, 8, 3, 4

  42. [50]

    Scorehypo: Probabilistic human mesh estimation with hypothesis scoring

    Yuan Xu, Xiaoxuan Ma, Jiajun Su, Wentao Zhu, Yu Qiao, and Yizhou Wang. Scorehypo: Probabilistic human mesh estimation with hypothesis scoring. In CVPR, pages 979–989,

  43. [51]

    Explicit box detection unifies end-to-end multi-person pose estimation

    Jie Yang, Ailing Zeng, Shilong Liu, Feng Li, Ruimao Zhang, and Lei Zhang. Explicit box detection unifies end-to-end multi-person pose estimation. In ICLR, 2023. 3

  44. [52]

    Capturing the motion of every joint: 3d human pose and shape estimation with independent tokens

    Sen Yang, Wen Heng, Gang Liu, Guozhong Luo, Wankou Yang, and YU Gang. Capturing the motion of every joint: 3d human pose and shape estimation with independent tokens. In ICLR, 2023. 3

  45. [53]

    Faster voxelpose: Real-time 3d human pose estimation by orthographic projection

    Hang Ye, Wentao Zhu, Chunyu Wang, Rujie Wu, and Yizhou Wang. Faster voxelpose: Real-time 3d human pose estimation by orthographic projection. In ECCV, 2022. 3

  46. [54]

    A-vit: Adaptive tokens for efficient vision transformer

    Hongxu Yin, Arash Vahdat, Jose M Alvarez, Arun Mallya, Jan Kautz, and Pavlo Molchanov. A-vit: Adaptive tokens for efficient vision transformer. In CVPR, pages 10809–10818,

  47. [55]

    Monocular 3d pose and shape estimation of multiple people in natural scenes-the importance of multiple scene constraints

    Andrei Zanfir, Elisabeta Marinoiu, and Cristian Sminchisescu. Monocular 3d pose and shape estimation of multiple people in natural scenes-the importance of multiple scene constraints. In CVPR, pages 2148–2157, 2018. 2, 3

  48. [56]

    Deep network for the in- tegrated 3d sensing of multiple people in natural images

    Andrei Zanfir, Elisabeta Marinoiu, Mihai Zanfir, Alin-Ionut Popa, and Cristian Sminchisescu. Deep network for the in- tegrated 3d sensing of multiple people in natural images. In NeurIPS, 2018. 2, 3

  49. [57]

    Pymaf: 3d human pose and shape regression with pyramidal mesh alignment feedback loop

    Hongwen Zhang, Yating Tian, Xinchi Zhou, Wanli Ouyang, Yebin Liu, Limin Wang, and Zhenan Sun. Pymaf: 3d human pose and shape regression with pyramidal mesh alignment feedback loop. In ICCV, pages 11446–11456, 2021. 3

  50. [58]

    Dino: Detr with improved denoising anchor boxes for end-to-end object de- tection

    Hao Zhang, Feng Li, Shilong Liu, Lei Zhang, Hang Su, Jun Zhu, Lionel Ni, and Heung-Yeung Shum. Dino: Detr with improved denoising anchor boxes for end-to-end object de- tection. In ICLR, 2023. 3 10

  51. [59]

    Motionbert: A unified perspective on learning human motion representations

    Wentao Zhu, Xiaoxuan Ma, Zhaoyang Liu, Libin Liu, Wayne Wu, and Yizhou Wang. Motionbert: A unified perspective on learning human motion representations. In CVPR, pages 15085–15099, 2023. 3

  52. [60]

    Social motion prediction with cognitive hierarchies

    Wentao Zhu, Jason Qin, Yuke Lou, Hang Ye, Xiaoxuan Ma, Hai Ci, and Yizhou Wang. Social motion prediction with cognitive hierarchies. In NeurIPS, 2023. 2

  53. [61]

    Human motion generation: A survey

    Wentao Zhu, Xiaoxuan Ma, Dongwoo Ro, Hai Ci, Jinlu Zhang, Jiaxin Shi, Feng Gao, Qi Tian, and Yizhou Wang. Human motion generation: A survey. IEEE TPAMI, 46(4): 2430–2449, 2024. 2 11 SAT-HMR: Real-Time Multi-Person 3D Mesh Estimation via Scale-Adaptive Tokens Supplementary Mate...

  54. [63]

    validation set, 3DPW [ 49] test set and CMU Panoptic

  55. [64]

    B3 (MuPoTS [29] is not included due to the lack of bounding box annotations)

    test set in Tab. B3 (MuPoTS [29] is not included due to the lack of bounding box annotations). To study the impact of scale-adaptive tokens, we evaluate our method (644*, mixed resolution) and the corresponding single-resolution baseline (644) on these datasets. As shown, AGOR...

  56. [65]

    However, 3DPW’s larger-scale instances show no improvement, likely due to the model’s limited capability

    and CMU Panoptic [15]. However, 3DPW’s larger-scale instances show no improvement, likely due to the model’s limited capability. Scale threshold αs. To further study the impact of high- resolution tokens, we conduct experiments on various scale thresholds αs while retaining th...

  57. [66]

    We report F1-Score (F1) and Mean Absolute Error (MAE) for evaluating scale map accuracy

    validation set. We report F1-Score (F1) and Mean Absolute Error (MAE) for evaluating scale map accuracy. To analyze its impact on mesh estimation, we replace the predicted scale map with GT and report MVE for different scale ranges and the average (Avg.). Scale F1↑ MAE↓ MVE↓ m...

  58. [2023]

    6, 7, 8, 1, 2, 3, 4, 5

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.