Pith. sign in

REVIEW 4 major objections 5 minor 31 references

EventEgoHands: Event-based Egocentric 3D Hand Mesh Reconstruction

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read EventEgoHands: masking background events makes egocentric event-based 3D hand mesh reconstruction feasible.

desk verdict First egocentric event-based hand mesh method with a segmentation module that clearly helps on their synthetic benchmark — but all evidence is v2e-simulated, so the real-world claim is unproven. read the letter →

arxiv 2505.19169 v3 pith:DG7F24UB submitted 2025-05-25 cs.CV

Ryosei Hara , Wataru Ikeda , Masashi Hatano , Mariko Isogawa This is my paper · ORCID
classification cs.CV
keywords 3DhandmeshreconstructioneventcameraegocentricvisionsegmentationpointcloudMANOmodelN-HOT3Ddatasetsimulation
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

EventEgoHands is offered as the first method to reconstruct a 3D hand mesh from an egocentric event camera, and its central claim is that background events from the wearer's motion, not a lack of hand signal, are what broke prior event-based hand tracking. The paper introduces a Hand Segmentation Module that predicts a hand mask from event frames, filters the event point cloud to the hand region, and then reconstructs both hands with a point-cloud network. On the synthetic N-HOT3D dataset, this raises R-AUC from 0.243/0.251 for two prior event-based methods to 0.450 and cuts mean joint error from about 106 mm to 59.51 mm and vertex error from about 65–70 mm to 42.92 mm. The practical stake is a route to hand tracking in low light and fast ego-motion for AR/VR and human-computer interaction, though all evidence comes from simulation.

What carries the argument

The load-bearing mechanism is a two-stage event filter: the Hand Segmentation Module takes T=3 Locally-Normalised Event Surfaces (a frame representation that weights events by recency within a time window) and a U-Net predicts a binary hand mask; the mask is then applied to the Event Cloud representation (raw events as 5D points with position, timestamp, and two polarity channels) to keep only hand events, downsampled to at most N=2048 points. The filtered cloud feeds the Hand Reconstruction Module, whose PointNet++ encoder extracts point features and cross-attention lets the left- and right-hand branches exchange information before a decoder outputs MANO parameters. The segmentation does two jobs at once: it removes the ego-motion background events that break prior methods, and it caps the point count so the cost scales with 2048 events rather than the full stream.

What would settle it

Record real egocentric event streams with an event camera and synchronized ground-truth hand meshes (for instance from motion-capture gloves or multi-view RGB), run the published EventEgoHands pipeline on them, and compare MPJPE to the reported 59.51 mm. If real-data error stays near the ~106 mm baseline level, the simulated-to-real transfer premise is the point of failure.

Watch

Extended reading notes

Core claim

The central discovery argued in the paper is that the dominant obstacle to egocentric event-camera hand reconstruction is not hand motion but background events created by the wearer's own movement, and that a learned mask can remove that obstacle. EventEgoHands first predicts a hand-region mask from three LNES event frames with a U-Net, then discards all event-cloud points outside the mask before a PointNet++ encoder with cross-attention between the two hand branches regresses MANO pose and shape parameters. On N-HOT3D, a synthetic dataset of 447,704 samples generated from HOT3D via v2e, the method reports R-AUC of 0.450 versus 0.243 and 0.251 for EventHands and Ev2Hands, MPJPE of 59.51 mm versus 105.80 and 106.75 mm, and MPVPE of 42.92 mm versus 65.23 and 69.66 mm. Ablations attribute the largest share of the gain to the Hand Segmentation Module: removing it drops R-AUC to 0.392 and raises MPJPE to 66.24 mm.

Load-bearing premise

The load-bearing premise is that v2e-simulated event streams from HOT3D RGB video faithfully reproduce real egocentric event-camera statistics, noise included, so that gains measured on N-HOT3D will transfer to physical event cameras.

Editorial extensions

If this is right

  • If the reported numbers hold, event-based egocentric hand mesh reconstruction can be done without RGB or depth input: the full pipeline uses only event frames and event clouds.
  • The mask-first design means computational cost scales with the masked event budget (N=2048) rather than the total event count, which makes real-time wearable use more plausible.
  • The roughly 43% reduction in mean joint error over EventHands and Ev2Hands indicates that background events, rather than hand-motion ambiguity, were the dominant error source in prior egocentric settings.
  • Because the method outputs MANO parameters directly, the reconstructed mesh can be used in AR/VR avatars or robotic teleoperation without an additional fitting step.

Reading between the lines

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

  • The paper does not test against real event hardware; the natural next check is whether the mask's relative gain survives on real streams with v2e-style noise removed, and absolute errors likely rise if it does not.
  • Since mask supervision comes from projected MANO ground truth, the design inherits MANO's limitations; the reported failure on object occlusion follows directly, and a hand-object interaction term is the obvious extension.
  • Because the segmentation ablation accounts for the largest performance gap, a cheaper non-learned baseline (e.g., fixed central region or motion-magnitude thresholding) would clarify how much of the gain is learned semantics versus coarse background rejection.
  • The same mask-then-reconstruct pattern is a plausible recipe for other event-based egocentric tasks degraded by ego-motion clutter, such as full-body pose or object tracking, but the paper presents no evidence for that transfer.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes EventEgoHands, the first method for event-based egocentric 3D hand mesh reconstruction. The method consists of two modules: a Hand Segmentation Module, which uses a U-Net on LNES event frames to predict a binary hand mask that filters out background events, and a Hand Reconstruction Module, which uses PointNet++ with Cross-Attention to regress MANO parameters for both hands from the filtered event cloud. To train and evaluate the method, the authors create N-HOT3D, a large synthetic dataset of 447,704 samples generated by applying the v2e event simulator to the HOT3D egocentric dataset. Experiments on N-HOT3D report substantial improvements over prior event-based hand reconstruction baselines: R-AUC increases from 0.243/0.251 (EventHands/Ev2Hands) to 0.450, MPJPE decreases from 105.80/106.75 mm to 59.51 mm, and MPVPE from 65.23/69.66 mm to 42.92 mm. Ablations show that removing the segmentation module reduces R-AUC to 0.392 and increases MPJPE to 66.24 mm, while replacing Cross-Attention with Self-Attention gives intermediate results.

Significance. If the reported results are robust, the paper makes a valuable contribution: it is the first to address egocentric event-based hand mesh reconstruction, and the proposed segmentation-based filtering is a simple and intuitively effective mechanism for suppressing camera-wearer-induced background events. The N-HOT3D dataset, though synthetic, is large and could be useful for future research. The improvements over adapted baselines are large and the ablation supports the central role of the segmentation module. However, the evidence is currently confined to a single synthetic dataset with no real-event-camera validation, no held-out mask quality metrics, and incomplete baseline adaptation details. These gaps limit the strength of the central claim that the method handles dynamic egocentric scenes in practice.

major comments (4)
  1. [§4.1 and §4.2] The dataset split is described inconsistently. Section 4.1 states that N-HOT3D is 'divided into 347,854 samples for training and 99,850 samples for evaluation,' while Section 4.2 states a split of approximately 278K (62%) for training, 70K (16%) for validation, and 99K (22%) for testing. These numbers do not match: the first split has no validation set, and the second split uses far fewer training samples. The discrepancy is load-bearing for reproducibility because the reported test results must be tied to a specific train/validation/test partition, and hyperparameter selection (Supplementary Tables 1 and 2) is presumably performed on a validation set. Please clarify the exact split and which split is used for the results in Table 1.
  2. [§4.1, §5.1, Table 1] The entire evaluation, including the stability of the segmentation module, rests on the synthetic N-HOT3D dataset generated with v2e. The central claim is that the Hand Segmentation Module mitigates background events caused by egocentric camera motion, but v2e's event generation model may not reproduce the statistics of real event sensors (e.g., DAVIS346) or the actual distribution of ego-motion-induced background events. The mask supervision itself is obtained by projecting HOT3D mesh ground truth onto event frames that are generated from distortion-corrected videos, so training and test are perfectly aligned with the same simulation pipeline. Without any evaluation on real event camera data (even qualitative), or at least a domain-gap analysis comparing event noise and background event statistics, the reported gains (e.g., R-AUC 0.450 vs. 0.392 without segmentation) may not transfer to physical event cameras. This is a load-bearing limitation for the practical claim of the paper.
  3. [§5.1 and Supplementary Table 1] No quantitative evaluation of the Hand Segmentation Module on the held-out test set is reported. The main paper reports only reconstruction metrics (R-AUC, MPJPE, MPVPE) and the ablation of removing the module. The supplementary Table 1 reports IoU only for hyperparameter selection of loss weights, not for the final model on the test set, and it is unclear whether that IoU is computed on a validation or test split. Because the segmentation module is a core contribution and its mechanism (filtering background events) is the explanatory hypothesis for the reconstruction improvement, the paper should report test-set mask IoU (and, ideally, the fraction of background events removed) to verify that the module indeed learns clean hand masks and that the improvement is not an artifact of, e.g., simply reducing the number of points in the cloud.
  4. [§4.3] The baseline adaptations are not fully specified, which makes the comparison potentially unfair. For Ev2Hands, the authors state that they 'first trained on the annotated Ev2Hands-S dataset and then fine-tuned it with N-HOT3D,' but they do not describe the fine-tuning objective, learning rate schedule, or how the absence of per-event left/right/background labels in N-HOT3D is handled during fine-tuning. Since Ev2Hands originally uses per-event classification as part of its pipeline, removing that supervision may significantly degrade it; the paper should specify the exact fine-tuning protocol. For EventHands, training separate left-hand and right-hand models is mentioned but the details (data cropping, joint outputs) are omitted. These omissions prevent readers from assessing whether the reported baseline numbers are reasonable.
minor comments (5)
  1. [§4.2] The number of epochs for both modules is fixed at 10, but no early stopping criterion or validation-based model selection is described; please clarify how the final model is chosen.
  2. [Supplementary §B.2] The fixed point-cloud size N=2048 is mentioned in the supplementary, but the main paper does not specify how the filtered event cloud of size N_f is converted to N=2048 points (e.g., random subsampling, padding, or selection of the most recent events). This matters for interpreting the computational cost reduction and the segmentation filter's effect.
  3. [Figure 3 caption] The phrase 'while independently providing the raw event and hand mask' is confusing; it likely means that the event stream and mask are generated from the RGB frame and MANO annotations, but the wording should be clarified.
  4. [Reference [20]] The reference to the DAVIS346 datasheet is written as 'iniV ation' with a stray space; please correct.
  5. [§5.1] All results are reported as single runs without error bars or multiple seeds. Given the large observed margins this is a minor concern, but adding variance estimates would strengthen the claims.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity found; the reported gains are measured on a held-out test split and are not forced by construction.

full rationale

The paper's central claim—that EventEgoHands outperforms EventHands and Ev2Hands on N-HOT3D (Table 1: R-AUC 0.450 vs 0.243/0.251; MPJPE 59.51 mm vs 105.80/106.75 mm)—is an empirical result on a test split held out from training, as stated in Section 4.2: 'splitting it into approximately 278K (62%) for training, 70K (16%) for validation, and 99K (22%) for testing.' The hand mask ground truth is produced by projecting the same MANO meshes that provide the reconstruction supervision (Section 4.1: 'we projected the provided 3D mesh ground-truth annotations onto the 2D space to create ground-truth hand masks'), which is standard supervision rather than a parameter fitted to the reported outputs. Loss weights (λμ, λ¶, λϵ, λ·) were selected by validation-set R-AUC/MPJPE/MPVPE (supplementary Table 2), and the reported numbers come from the held-out test split, so no fitted input is being presented as a prediction. The method adopts LNES [11] and Event Cloud [13] representations from prior work by other groups, and there is no load-bearing self-citation chain; the cited baselines are retrained on the same data for fair comparison. The only caveat is that N-HOT3D is a synthetic dataset generated by the authors via the v2e simulator from HOT3D, so the evaluation is not fully external to the authors; also, the paper does not report mask IoU on the held-out test set or the fraction of background events removed, which limits verification of the segmentation mechanism. These are evidence and generalizability gaps that belong to correctness risk, not circularity: the test-set predictions are not constructed from the reported metrics or from a fit to the test labels.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central claim depends mainly on the realism of the synthetic event simulation and on the choice of loss weighting and event subsampling; these are the quantities the reader must take on faith because they are not externally benchmarked.

free parameters (4)
  • Hand Reconstruction Module loss weights (lambda_mu, lambda_nu, lambda_epsilon, lambda_tau) = 0.1, 1.0, 1.0, 20
    Chosen via one-at-a-time ablation in Supplementary Table 2; the central results depend on these values, and neighboring values change R-AUC by up to 0.03 and MPJPE by several mm.
  • Hand Segmentation Module loss weights (lambda_BCE, lambda_Dice) = 0.7, 0.3
    Selected from three candidates in Supplementary Table 1 based on mask IoU; IoU ranges from 0.519 to 0.567 across the sweep.
  • Number of event points per cloud N = 2048
    Chosen to cap computational cost; the dataset contains up to 100K events per window, and N=2048 is a subsampling level that trades information against training time (Supplementary Section C).
  • Number of time windows T = 3
    Adopted to supply temporal context to the segmentation module; the paper states T=3 corresponds to 3 frames at 30 fps.
assumptions (4)
  • domain assumption The v2e event simulator produces event streams that adequately represent real egocentric event camera data for training and evaluation.
    Invoked in Section 4.1 where N-HOT3D is generated from HOT3D RGB videos with v2e; the paper does not verify against real event camera recordings.
  • domain assumption The ground-truth MANO parameters and 3D annotations in HOT3D are accurate, and projecting the mesh to 2D yields correct hand segmentation masks.
    Used in Section 4.1 to create ground-truth hand masks; if the HOT3D annotations contain errors, both segmentation supervision and reconstruction target are affected.
  • domain assumption The mask predicted for the latest time step is valid for filtering event clouds across the entire T=3 window.
    Stated in Supplementary Section B.1: 'the mask at the latest timestamp is used as the filtering mask'; this assumes the hand region does not move significantly within the window.
  • standard math U-Net, PointNet++, and cross-attention architectures as borrowed from prior work are appropriate inductive biases for this task.
    The paper adopts these components from references [17], [18], and [13]; no new theoretical guarantees are provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of EventEgoHands: Event-based Egocentric 3D Hand Mesh Reconstruction." pith.science (2026). https://pith.science/paper/DG7F24UB

@misc{pith2026250519169,
  author       = {Pith},
  title        = {Pith review of: EventEgoHands: Event-based Egocentric 3D Hand Mesh Reconstruction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DG7F24UB}},
  note         = {Machine review of arXiv:2505.19169}
}
read the original abstract

Reconstructing 3D hand mesh is challenging but an important task for human-computer interaction and AR/VR applications. In particular, RGB and/or depth cameras have been widely used in this task. However, methods using these conventional cameras face challenges in low-light environments and during motion blur. Thus, to address these limitations, event cameras have been attracting attention in recent years for their high dynamic range and high temporal resolution. Despite their advantages, event cameras are sensitive to background noise or camera motion, which has limited existing studies to static backgrounds and fixed cameras. In this study, we propose EventEgoHands, a novel method for event-based 3D hand mesh reconstruction in an egocentric view. Our approach introduces a Hand Segmentation Module that extracts hand regions, effectively mitigating the influence of dynamic background events. We evaluated our approach and demonstrated its effectiveness on the N-HOT3D dataset, improving MPJPE by approximately more than 4.5 cm (43%).

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 31 canonical work pages

  1. [1]

    EventEgoHands: Event-based Egocentric 3D Hand Mesh Reconstruction

    INTRODUCTION 3D human hand mesh reconstruction has many applica- tions, and many methods have been proposed up to now [ 1, 2, 3, 4, 5]. In particular, methods using egocentric cameras [ 1, 3, 5] are essential for applications such as AR/VR, and robotics. However, all of these existing methods rely on RGB(D)- based approaches and they have two main challen...

    work page Pith review arXiv 2025

  2. [2]

    3D Hand Mesh Reconstruction 3D hand mesh reconstruction has seen significant ad- vancements, most of which rely on RGB [ 1, 2, 3] or depth [ 4, 5] sensors

    RELATED WORKS 2.1. 3D Hand Mesh Reconstruction 3D hand mesh reconstruction has seen significant ad- vancements, most of which rely on RGB [ 1, 2, 3] or depth [ 4, 5] sensors. Parametric models, such as MANO [ 16], are often used to represent hand meshes for reconstruction. In this study, we also utilized MANO parameters provided by the HOT3D [ 15] dataset....

    work page

  3. [3]

    As shown in Fig

    PROPOSED METHOD We propose EventEgoHands, a 3D hand mesh reconstruc- tion method that uses only event data in dynamic scenes from an egocentric view. As shown in Fig. 2, given a sequence of event frame I 1:T and event point clouds E1:T segmented into T fixed-width time windows, EventEgoHands reconstructs 3D joint positions J ∈ R20× 3 and mesh vertex positi...

    work page

  4. [4]

    2 cm (34%) compared to previous methods

    5 cm (43%) and MPVPE by more than 2. 2 cm (34%) compared to previous methods

    work page

  5. [5]

    Dataset To train our model, we needed event data with ground- truth annotations of both hands from an egocentric view

    EXPERIMENT 4.1. Dataset To train our model, we needed event data with ground- truth annotations of both hands from an egocentric view. However, there are no datasets from an egocentric view. Thus, we created an event dataset N-HOT3D using the event simulator v2e [ 14] from the HOT3D [ 15] dataset. We specifi- cally used a subset of the Aria glasses data wi...

    work page

  6. [6]

    Using these values, we conducted distortion correction on the videos and then converted the corrected videos using the event simulator

    We used MANO parameters, camera extrinsic param- eters, and intrinsic parameters given by HOT3D. Using these values, we conducted distortion correction on the videos and then converted the corrected videos using the event simulator. The output events size was set to 346 × 260, based on the resolution of the DA VIS346 [ 20] event camera, which is also used...

    work page

  7. [7]

    Quantitative Evaluation To evaluate the effectiveness of our proposed method, we conducted a quantitative evaluation with baseline methods; the results are shown in Table

    RESULTS 5.1. Quantitative Evaluation To evaluate the effectiveness of our proposed method, we conducted a quantitative evaluation with baseline methods; the results are shown in Table

    work page

  8. [8]

    In particular, our method improved the R-AUC by approximately 0.2 points compared to the baseline

    The results demonstrate that the proposed method outperforms all baselines. In particular, our method improved the R-AUC by approximately 0.2 points compared to the baseline. Additionally, MPJPE improved by about 45 mm and MPVPE by about 20 mm. To further analyze the effectiveness of the Hand Segmen- tation Module and Cross-Attention, we conducted an abla...

    work page

Show all 31 references
  1. [9]

    CONCLUSION In this study, we proposed EventEgoHands, the first method of event-based egocentric 3D hand mesh reconstruc- tion. Our Hand Segmentation Module extracts only events in the hand region, thereby reducing the influence of back- ground events caused by the motion of the ...

  2. [10]

    3D hand pose estimation in everyday egocentric images,

    Aditya Prakash, Ruisen Tu, Matthew Chang, and Saurabh Gupta, “3D hand pose estimation in everyday egocentric images,” in ECCV, 2024, pp. 183–202

    2024

  3. [11]

    Reconstructing hands in 3D with transformers,

    Georgios Pavlakos, Dandan Shan, Ilija Radosavovic, Angjoo Kanazawa, David Fouhey, and Jitendra Malik, “Reconstructing hands in 3D with transformers,” in CVPR, 2024, pp. 9826–9836

    2024

  4. [12]

    AssemblyHands: towards egocentric activity understanding via 3D hand pose estimation,

    Takehiko Ohkawa, Kun He, Fadime Sener, Tomas Ho- dan, Luan Tran, and Cem Keskin, “AssemblyHands: towards egocentric activity understanding via 3D hand pose estimation,” in CVPR, 2023, pp. 12999–13008

    2023

  5. [13]

    HandDAGT: a denoising adaptive graph transformer for 3D hand pose estimation,

    Wencan Cheng, Eunji Kim, and Jong Hwan Ko, “HandDAGT: a denoising adaptive graph transformer for 3D hand pose estimation,” in ECCV, 2024, pp. 35– 52

    2024

  6. [14]

    Real-time hand tracking under occlusion from an egocentric rgb-d sensor,

    Franziska Mueller, Dushyant Mehta, Oleksandr Sot- nychenko, Srinath Sridhar, Dan Casas, and Christian Theobalt, “Real-time hand tracking under occlusion from an egocentric rgb-d sensor,” in ICCV, 2017, pp. 1163–1172

    2017

  7. [15]

    Event-Based Vi- sion: A Survey,

    Guillermo Gallego, Tobi Delbruck, Garrick Orchard, Chiara Bartolozzi, Brian Taba, Andrea Censi, Stefan Leutenegger, Andrew J. Davison, Jorg Conradt, Kostas Daniilidis, and Davide Scaramuzza, “Event-Based Vi- sion: A Survey,” TPAMI, vol. 44, no. 01, pp. 154–180, 2022

    2022

  8. [16]

    Secrets of event-based optical flow, depth and ego-motion estimation by contrast maximiza- tion,

    Shintaro Shiba, Y annick Klose, Y oshimitsu Aoki, and Guillermo Gallego, “Secrets of event-based optical flow, depth and ego-motion estimation by contrast maximiza- tion,” TPAMI, vol. 46, no. 12, pp. 7742–7759, 2024

    2024

  9. [17]

    Event-based se- mantic segmentation with posterior attention,

    Zexi Jia, Kaichao Y ou, Weihua He, Y ang Tian, Y ongx- iang Feng, Y aoyuan Wang, Xu Jia, Yihang Lou, Jingyi Zhang, Guoqi Li, and Ziyang Zhang, “Event-based se- mantic segmentation with posterior attention,” IEEE Transactions on Image Processing , vol. 32, pp. 1829– 1842, 2023

    2023

  10. [18]

    EventEgo3D: 3D human motion capture from egocentric event streams,

    Christen Millerdurai, Hiroyasu Akada, Jian Wang, Diogo Luvizon, Christian Theobalt, and Vladislav Golyanik, “EventEgo3D: 3D human motion capture from egocentric event streams,” in CVPR, 2024, pp. 1186–1195

    2024

  11. [19]

    EventHPE: event-based 3D human pose and shape es- timation,

    Shihao Zou, Chuan Guo, Xinxin Zuo, Sen Wang, Hu Xi- aoqin, Shoushun Chen, Minglun Gong, and Li Cheng, “EventHPE: event-based 3D human pose and shape es- timation,” in ICCV, 2021, pp. 10996–11005

    2021

  12. [20]

    EventHands: real-time neu- ral 3D hand pose estimation from an event stream,

    Viktor Rudnev, Vladislav Golyanik, Jiayi Wang, Hans- Peter Seidel, Franziska Mueller, Mohamed Elgharib, and Christian Theobalt, “EventHands: real-time neu- ral 3D hand pose estimation from an event stream,” in ICCV, 2021, pp. 12385–12395

    2021

  13. [21]

    EvHandPose: event-based 3D hand pose estimation with sparse supervision,

    Jianping Jiang, Jiahe Li, Baowen Zhang, Xiaoming Deng, and Boxin Shi, “EvHandPose: event-based 3D hand pose estimation with sparse supervision,” TPAMI, vol. 46, no. 9, pp. 6416–6430, 2024

    2024

  14. [23]

    v2e: From video frames to realistic DVS events,

    Y uhuang Hu, Shih-Chii Liu, and Tobi Delbruck, “v2e: From video frames to realistic DVS events,” in CVPRW, 2021, pp. 1312–1321

    2021

  15. [24]

    HOT3D: hand and object track- ing in 3D from egocentric multi-view videos,

    Prithviraj Banerjee, Sindi Shkodrani, Pierre Moulon, Shreyas Hampali, Shangchen Han, Fan Zhang, Lin- guang Zhang, Jade Fountain, Edward Miller, Selen Ba- sol, Richard Newcombe, Robert Wang, Jakob Julian En- gel, and Tomas Hodan, “HOT3D: hand and object track- ing in 3D from eg...

    2024

  16. [25]

    Embodied Hands: modeling and capturing hands and bodies together,

    Javier Romero, Dimitrios Tzionas, and Michael J. Black, “Embodied Hands: modeling and capturing hands and bodies together,” ACM Transactions on Graphics, vol. 36, no. 6, 2017

    2017

  17. [26]

    U-Net: convolutional networks for biomedical image segmentation,

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox, “U-Net: convolutional networks for biomedical image segmentation,” in MICCAI, 2015, pp. 234–241

    2015

  18. [27]

    PointNet++: deep hierarchical feature learning on point sets in a metric space,

    Charles R. Qi, Li Yi, Hao Su, and Leonidas J. Guibas, “PointNet++: deep hierarchical feature learning on point sets in a metric space,” in NeurIPS, 2017, p. 5105–5114

    2017

  19. [28]

    Generalised dice overlap as a deep learning loss function for highly un- balanced segmentations,

    Carole H. Sudre, Wenqi Li, Tom V ercauteren, Sebastien Ourselin, and M. Jorge Cardoso, “Generalised dice overlap as a deep learning loss function for highly un- balanced segmentations,” in Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Su...

    2017

  20. [29]

    Davis346,

    iniV ation, “Davis346,” https://inivation. com/wp-content/uploads/2019/08/ DAVIS346.pdf , 2019

    2019

  21. [30]

    Adam: a method for stochastic optimization,

    Diederik P . Kingma and Jimmy Ba, “Adam: a method for stochastic optimization,” in ICLR, 2015. EVENTEGOHANDS: EVENT-BASED EGOCENTRIC 3D HAND MESH RECONSTRUCTION Supplementary Material Contents A Overview of the Supplementary Material 1 B Hand Segmentation Module 1 B.1 Locally-...

    2015

  22. [31]

    EventHands: real-time neural 3D hand pose estimation from an event stream,

    Viktor Rudnev, Vladislav Golyanik, Jiayi Wang, Hans- Peter Seidel, Franziska Mueller, Mohamed Elgharib, and Christian Theobalt, “EventHands: real-time neural 3D hand pose estimation from an event stream,” in ICCV, 2021, pp. 12385–12395

    2021

  23. [32]

    3D pose estimation of two inter- acting hands from a monocular event camera,

    Christen Millerdurai, Diogo Luvizon, Viktor Rudnev, Andr´e Jonas, Jiayi Wang, Christian Theobalt, and Vladislav Golyanik, “3D pose estimation of two inter- acting hands from a monocular event camera,” in 3DV, 2024, pp. 291–301

    2024

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.