Pith. sign in

REVIEW 4 major objections 4 minor 68 references

Marrying Optimal Transport and ODEs for Unified Continuous-Time 4D Reconstruction and Tracking

T0 review · 4 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Uni4R claims that a single continuous velocity field, built from an Optimal Transport straight-line path on anchor features and integrated by an ODE, resolves the ill-posed continuous inverse problem and unifies 4D reconstruction with…

desk verdict Useful empirical system for continuous-time 4D tracking, but the OT kinematic prior is asserted more than proven. read the letter →

arxiv 2608.09613 v1 pith:SCGK4UGH submitted 2026-08-10 cs.CV

Liying Yang , Hao Mo , Jialun Liu , Chen Liu , Xinxing Yu , Chenhao Guan , Hui Ma , Xiao Cao
show 2 more authors
Ajian Liu Yanyan Liang
This is my paper · ORCID
classification cs.CV MSC 68T4549Q2265L05
keywords continuous4Dreconstructionpointtrackingvelocityfieldoptimaltransportflowmatchingordinarydifferentialequationintegral-consistencytrainingkinematics-awarebenchmark
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 sets out to prove that 4D reconstruction and point tracking are the same problem when motion is described by a continuous velocity field, and that such a field can be learned from ordinary integer-frame supervision. It proposes Uni4R, a feed-forward transformer whose Flow Matching Guided Decoder uses Optimal Transport to define a straight probability path on anchor features, giving the network a kinematic prior for the direction of motion, and an ODE solver to integrate velocities to any query time. Because ground-truth velocities at fractional timestamps do not exist, the paper adds an integral-consistency training loss that integrates velocities forward and compares the result with the next integer frame's pointmap, so the velocity field is supervised end-to-end from data that does exist. The payoff claimed is continuous-time kinematic coherence: smooth trajectories, preserved local geometry at fractional frames, and improved accuracy at integer frames as well, with state-of-the-art results on four tracking datasets, two reconstruction datasets, and the paper's new kinematics-aware benchmark.

What carries the argument

The load-bearing object is the FM-guided velocity feature $X_{\mathrm{FM}}^\tau = \tau X_1 + (1-\tau) X_2$, a convex combination of anchor features extracted by the global velocity branch at integer frames; Flow Matching theory says this OT path is the minimum-kinetic-energy straight line between source and target features. That feature is used as the query in a spatial-temporal cross-attention module that pulls geometric context from the point reconstruction branch, and the conditional flow matching loss $L_{\mathrm{CFM}}$ regresses predicted velocity features onto the feature difference $X_1-X_2$. Around this object, the paper builds an anchor-based ODE that integrates velocity only between nearest integer anchors to avoid drift, a local velocity prediction module that injects sinusoidal time embeddings, and an integral-consistency training loss $L_{\mathrm{ICT}}$ that supervises the continuous field from integer pointmaps.

What would settle it

Take a synthetic sequence with known strong acceleration, such as a ball under gravity or a car braking, query the trained Uni4R at fractional timestamps, and compare its predicted pointmaps and trajectory derivatives against dense ground-truth. If the OT straight-line feature path is genuinely kinematic, fractional predictions should stay closer to the ground truth than the B-spline baseline and the $L_{\mathrm{ICT}}$-only variant; if the straight-line prior smooths away acceleration, the predicted fractional geometry will lag or lead the true positions, with MAM and ELV errors rising with acceleration. A shorter decisive check is to retrain the model with a deliberately curved feature path, such as quadratic interpolation of anchor features; if the kinematics metrics do not worsen, the straight-line OT prior is not the mechanism producing the reported gains.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is a way to make the inverse problem of recovering continuous motion from discrete frames well-posed. Uni4R learns a velocity field $V(P(t),t)$ such that a point's position at any time $k+\tau$ is obtained by the anchor-based ODE $P(k+\tau)=P(k)+\int_k^{k+\tau} V(P(t),t)\,dt$. The ill-posedness, which is that infinitely many velocity fields fit the integer frames, is broken by the Flow Matching Guided Decoder, which defines an Optimal Transport straight-line probability path on an anchor feature manifold, $X_{\mathrm{FM}}^\tau = \tau X_1 + (1-\tau) X_2$, and trains the velocity decoder with the conditional flow matching loss $L_{\mathrm{CFM}}$ against the feature difference $X_1-X_2$. This is claimed to make the global direction of velocity kinematically meaningful. A second ingredient, the integral-consistency loss $L_{\mathrm{ICT}}$, integrates predicted velocities through an Euler ODE solver and supervises the result with the ground-truth pointmap at the next integer frame, so no fractional-frame velocity labels are needed. The paper reports state-of-the-art average percentage of points within distance for 3D point tracking and 4D reconstruction, and the lowest mean acceleration magnitude, edge-length variance, and local pairwise-distance variance on its continuous-time kinematics benchmark.

Load-bearing premise

The paper assumes that a straight line between anchor features in learned feature space, $X_{\mathrm{FM}}^\tau = \tau X_1 + (1-\tau) X_2$, points along the same direction as true 3D motion; if feature-space linearity does not correspond to physically meaningful point velocities, the kinematic prior that resolves the ill-posed inverse problem fails.

Editorial extensions

If this is right

  • At any fractional timestamp $\tau$, point positions are obtained by ODE integration from the nearest integer anchor, so the model can be queried continuously rather than limited to discrete frames.
  • The same velocity field drives both geometry and correspondences: reconstruction provides geometric features for velocity decoding, and tracking provides kinematic guidance for deformation; joint training outperforms training either branch alone.
  • Because the OT straight-line path reduces discretization error, a first-order Euler solver with step size 0.1 is sufficient; Uni4R runs in 1.57 seconds on 80-frame inputs, faster than the B-spline-based TraceAnything.
  • Removing any component of the framework, whether the flow matching loss, the FM-guided features, the integral-consistency training, or the local velocity prediction module, degrades both tracking and reconstruction, supporting the claim that each is load-bearing.
  • The kinematics-aware benchmark supplies dense fractional-frame ground truth and metrics such as MAM, ELV, VLPD, and APD, so continuous-time coherence can be measured rather than assumed.

Reading between the lines

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

  • If the paper is right that the OT straight-line feature path is the kinematic prior, the same construction should transfer to other continuous spatiotemporal outputs, such as scene flow, fractional-time optical flow, or dynamic point-cloud interpolation, wherever velocity labels are absent; a cheap test is to swap the regression target from pointmap differences to scene-flow labels and see whether
  • The integral-consistency loss is a general device for ODE-based predictors with coarse-time supervision: integrate to the next observation and backpropagate. A direct test would be to train the same network with a curved, non-OT feature path; if the kinematic metrics do not worsen, then the straight-line prior is not carrying the claimed load.
  • The benchmark's dense fractional ground truth allows a sharper kinematic audit than the paper reports: compare predicted velocities at fractional timestamps with finite differences of ground-truth trajectories. If the learned field deviates systematically under acceleration, the prior is enforcing smoothness rather than physics.
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

4 major / 4 minor

Summary. The paper proposes Uni4R, a unified framework for continuous-time 4D reconstruction and point tracking. It models dynamics via an anchor-based ODE and introduces a Flow Matching Guided Decoder (FMGD) that uses optimal-transport-inspired linear interpolation on learned anchor features as a kinematic prior. An integral-consistency training loss supervises the velocity field from integer-frame pointmaps using an ODE solver, and a new kinematics-aware benchmark is introduced. The paper reports state-of-the-art results on Point Odyssey, Dynamic Replica, Aerial Digital Twin, Panoptic Studio, and TUM-Dynamics, along with ablations that show the contribution of each component.

Significance. If the central mechanism were verified, the combination of flow matching and ODEs for continuous-time dynamic 3D vision would be a novel and potentially impactful contribution. The paper also introduces a dedicated benchmark for kinematic coherence at fractional timestamps, which could be a useful community resource. The empirical study is internally consistent, the ablations in Tables 5 and 6 isolate individual components, and the reported efficiency gain over Trace Anything is notable. However, the core claim that the OT straight-line path on anchor features provides a physically meaningful kinematic prior is not supported by direct evidence, which limits the significance of the theoretical contribution as currently presented.

major comments (4)
  1. [Sec. 3.2, Eq. (2)] The central claim that FMGD "resolves the ill-posed continuous inverse problem" (Sec. 3.2) rests on the assumption that the OT straight-line path on anchor features, X_FM^τ = τ X1 + (1−τ) X2, induces a kinematic prior on the physical 3D velocity field V(P(t),t) used in the ODE of Eq. (1). However, the conditional flow matching loss in Eq. (2) is computed entirely on the learned anchor features X, which are produced by the same network that also produces F_vel. Nothing in the formulation or in the experiments links feature-space linearity to minimum-energy or physically meaningful 3D motion. The ablations in Table 5 (w/o L_CFM and w/o FVF) show that removing these components degrades performance, but they do not isolate the OT straight-line assumption from the additional conditioning and architectural changes. An experiment comparing the proposed interpolation against, for example, a learned or random feature interpolation on a dataset with ground-truth instantaneous velocities would be needed to support the claimed mechanism.
  2. [Sec. 3.3, Eq. (3)] The integral-consistency loss (Eq. (3)) enforces only that the integrated trajectory ends at the ground-truth pointmap at T+1. As the paper itself notes (Sec. 3.1), infinitely many continuous velocity fields satisfy such an endpoint constraint; the ability of L_ICT to select a kinematically correct field therefore depends entirely on the prior induced by L_CFM. Since the prior is unverified (see previous comment), the claim that Uni4R yields kinematically coherent trajectories is not established. The kinematic metrics in Table 4 (MAM, ELV, VLPD) measure temporal smoothness and local rigidity, which can also be satisfied by smooth but incorrect trajectories. The paper should evaluate fractional-time predictions against ground-truth instantaneous positions or velocities, e.g., on the proposed Kubric-based benchmark, and report such errors separately.
  3. [Sec. 4.1, Table 1] The paper reports single-run results without error bars or multiple seeds. The margin over V-DPM on Point Odyssey All Points is 0.32 APD (81.59 vs 81.27), which is likely within run-to-run variation. Similar caveats apply to other close numbers, e.g., in Table 4's APD3D. Without additional runs or statistical significance measures, the "SOTA" claims are not robust. The authors should provide at least 3 seeds with mean and variance for the main comparisons.
  4. [Sec. 3.3, 'Kinematics-Aware Benchmark'] The new benchmark is a key contribution, but it is described only briefly and most details are postponed to an appendix that is not included in the manuscript. To allow reproducibility and to assess the validity of the metrics, the paper should specify the scene generation process, the number of sequences, the exact definitions of MAM, ELV, VLPD and APD, and the evaluation protocol (including how fractional ground-truth is obtained and how baselines are evaluated). Additionally, the paper should show that the proposed metrics are correlated with actual physical velocity errors, rather than merely with smoothness.
minor comments (4)
  1. [Sec. 3.2] The phrase "the local velocity prediction module then joint above features" should be "joins the above features".
  2. [Sec. 4.2, Table 2] The grouping "w/ Global Align" and "Feedforward" is confusing because the text states all methods use global median scaling; please clarify the alignment protocol for each group.
  3. [Sec. 4.3, Table 4] The baseline descriptions are unclear; specify what each "Baseline w/ ..." variant trains (e.g., whether L_ICT and L_CFM are used) and how the MLP predicts velocities.
  4. [Throughout] Several equations and expressions have rendering issues (e.g., Eq. (3) appears garbled in the manuscript); these should be corrected in the camera-ready version.

Circularity Check

2 steps flagged · score 6.0 of 10

FMGD's 'kinematic prior' is a linear interpolation of the network's own anchor features; the velocity head is trained to regress that same feature difference, so the continuous-time 'velocity prediction' is partially a self-fitted quantity relabeled as physical kinematics.

  1. fitted input called prediction [Sec. 3.2, final paragraph of 'Flow Matching Guided Feature Formulation' (after Eq. (2))]
    "Finally, the LVP outputs velocity field features Fvel ∈ R^{B×L×T×N×D}. We perform the CFM loss between X1 − X2 (as the value of ut(·)) and Fvel (as the value of vθ(·)) via Eq. (2). Moreover, convolution layers decode Fvel into continuous velocity field V."

    The CFM regression target u_t = X1 − X2 is a difference of anchor features extracted by the paper's own global velocity branch (X1 = {X_2,...,X_{T+1}}, X2 = {X_1,...,X_T}). Training Fvel, and hence the decoded velocity field V, to match this target is a regression onto a quantity the network itself constructs. The paper then calls this a 'robust kinematic inductive bias' and claims it resolves the ill-posed inverse problem. But without ground-truth fractional velocities, this loss can be minimized by reshaping the anchor features; the only external velocity-related constraint is L_ICT (Eq. 3), which fixes only the integral of V over [T, T+1], not the pointwise velocity field.

  2. ansatz smuggled in via citation [Sec. 3.2, 'Flow Matching Guided Feature Formulation' (first paragraph)]
    "According to the Flow Matching theorem for OT, the efficient probability path between a source state x0 and a target state x1 is a straight line [7]. For any arbitrary time t, we define the target conditional latent velocity field ut to construct this optimal path as ut(x|x0, x1) = x1 − x0. This formulation minimizes kinetic energy [54] and significantly reduces numerical truncation errors during continuous integration [55]."

    The cited OT straight-line and minimum-kinetic-energy results (Lipman et al. [7]; Benamou-Brenier [54]) are theorems about probability paths in the space where x0 and x1 live, here the learned anchor-feature manifold. The ODE in Eq. (1) is a physical 3D trajectory, and the paper provides neither proof nor experiment that straightness in feature space corresponds to minimum-kinetic-energy or true 3D point motion. Invoking these citations to justify a 'kinematic inductive bias' thus imports an ansatz as if it were a theorem: the claimed OT-based prior is exactly the construction X_FM^τ = τX1 + (1−τ)X2 that the method itself inserts, rather than a result derived from or validated against physical velocity data.

full rationale

Uni4R is not a pure tautology: L_ICT (Eq. 3) integrates the predicted velocity field and compares against ground-truth pointmaps at integer timestamps, L_DUSt3R provides standard geometric supervision, and Tables 1 and 2 report gains on external benchmarks (PO, DR, ADT, PStudio, TUM-Dynamics) using official baseline code. There is no load-bearing self-citation chain, and the method has independent empirical content, so the paper is not a 8-10 definitional or self-citation collapse. However, the central claim that FMGD resolves the ill-posed continuous inverse problem through an OT kinematic prior has a partial circular component. Eq. (2)'s target u_t = X1 − X2 is not an externally measured velocity; it is the difference of anchor features produced by the same network, and Fvel is trained to reproduce it, with the decoded V inheriting this self-referential objective. The OT straight-line/min-energy theorems from [7] and [54] are imported from generative modeling and are not shown to transfer from the anchor-feature manifold to the physical 3D ODE in Eq. (1). Since L_ICT is only an endpoint integral constraint, infinitely many velocity fields satisfy it, so the uniqueness and kinematic-coherence claims are carried by the feature-interpolation ansatz itself rather than by an independent velocity signal. The remaining supervision comes from real pointmaps, which keeps the overall contribution partially empirical rather than fully circular. Overall: partial circularity at the level of the velocity-prediction target and the kinematic-prior justification, warranting a score of 6 rather than 0 or 10.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The paper's theoretical grounding rests on the Flow Matching theorem and OT results from prior literature, plus several domain assumptions about feature manifolds and anchor quality. No machine-checked proofs, no code, no independent entity.

free parameters (2)
  • Integration step size h = 0.1
    Chosen for inference/evaluation; ablation study deferred to Appendix; affects accuracy vs. runtime trade-off.
  • Loss weights lambda1/lambda2/lambda3 = 1:1:1
    Set equal by hand without sensitivity analysis.
assumptions (5)
  • standard math OT straight-line path is the optimal (minimal kinetic energy) path between states x0 and x1 in the feature manifold.
    Used in Eq. (2) to define the regression target u_t = x1 - x0, citing [7,53,54]. Standard result in optimal transport theory, but its transfer to learned feature manifolds is an assumption.
  • domain assumption Anchor features X at integer timestamps capture the global dynamic state and implicitly encode motion information for fractional timestamps.
    Assumed in Sec 3.2 without proof; the entire FMGD construction depends on anchor features carrying sufficient kinematic information.
  • ad hoc to paper Euler integration with step size 0.1 yields accurate trajectories because the OT-based velocity field reduces discretization error.
    Claimed in Sec 3.3 and 4; no error-bound analysis; step size is hand-chosen.
  • domain assumption The pointmap predictions at integer timestamps (used as integration anchors and for supervision) are accurate enough to train the velocity field.
    Integral-consistency loss in Eq. (3) assumes P_T and P_{T+1} are reliable; errors in the DPT head directly propagate through the ODE path.
  • ad hoc to paper Spatial-temporal cross-attention with FM-guided queries enforces geometric and temporal smoothness.
    Design choice in Sec 3.2 (Fig. 3); no formal guarantee, only ablation C-F supports it empirically.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Marrying Optimal Transport and ODEs for Unified Continuous-Time 4D Reconstruction and Tracking." pith.science (2026). https://pith.science/paper/SCGK4UGH

@misc{pith2026260809613,
  author       = {Pith},
  title        = {Pith review of: Marrying Optimal Transport and ODEs for Unified Continuous-Time 4D Reconstruction and Tracking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SCGK4UGH}},
  note         = {Machine review of arXiv:2608.09613}
}
read the original abstract

Existing unified 4D reconstruction and point tracking approaches typically rely on heuristic interpolations or just predict at integer timestamps, lacking kinematic coherence and failing to model dynamics at any arbitrary timestamp. In this paper, we propose Uni4R, a framework that unifies these tasks by learning continuous velocity fields through the synergy of Optimal Transport (OT) and Ordinary Differential Equation (ODE). Importantly, this continuous velocity field acts as a kinematic prior that mutually benefits both 4D reconstruction and point tracking. Specifically, we propose the Flow Matching Guided Decoder (FMGD). A global velocity branch first extracts anchor features that capture the global dynamic state of the sequence. Then, FMGD leverages Flow Matching (FM) theory to formulate a probability path defined by OT on the anchor feature manifold, instantiating it as FM-guided velocity features for velocity prediction. This establishes a robust kinematic inductive bias. Meanwhile, a point reconstruction branch provides geometric features. The local velocity prediction module then joint above features and time embeddings, to decode velocities at arbitrary timestamps. To overcome the absence of high-quality ground-truth velocities in fractional frames, we propose an integral-consistency training strategy. This strategy uses an ODE solver to integrate velocities to recover target pointmaps, enabling the model to be supervised end-to-end directly from integer timestamps. Experimental results demonstrate that Uni4R achieves SOTA performance in both 4D reconstruction and point tracking, and achieves SOTA in our new kinematics-aware benchmark at continuous time.

Figures

Figures reproduced from arXiv: 2608.09613 by the authors.

Figure 1
Figure 1. Left: Uni4R is a feed-forward transformer that queries point maps at arbitrary timestamp. Right: While existing methods like TraceAnything [3] rely on B-spline fitting and inherently lack kinematic coherence, Uni4R learns continuous velocity fields via the synergy of OT and ODE. Specifically, we formulate a probability path defined by OT on the anchor feature manifold to establish a kinematic inductive bias. Then OD… view at source ↗
Figure 2
Figure 2. Overview of Uni4R. Given a sequence of input frames, a ViT encoder extracts frame features, which are subsequently decoded into velocity (anchor) features X and geometric features F R by global velocity and point reconstruction branches. Then we formulate Flow Matching (FM)-guided velocity features FV constructed by an probability path defined by OT on the anchor feature manifold. A Local Velocity Prediction (LVP) m… view at source ↗
Figure 3
Figure 3. The architecture of LVP. Flow Matching Guided Feature Formulation: According to the Flow Matching theorem for OT, the efficient probability path between a source state x0 and a target state x1 is a straight line [7]. For any arbitrary time t, we define the target conditional latent velocity field ut to construct this optimal path as ut(x|x0, x1) = x1 − x0. This formulation minimizes kinetic energy [54] and significa… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Comparison with Continuous Modeling Variant and Trace Anything [3] at fractional times￾tamp. We visualize the dynamic part. With integer time frames as input, baselines produce unstable geom￾etry structures with distorted windows and wheels at fractional timestep, our …
Figure 5
Figure 5. Figure 5: Qualitative Results of 4D Reconstruction and Point Tracking on real-world dataset DAVIS [65]. We visualize our Uni4R and baselines, including St4RTrack [1] (we apply their test-time adaptation), Trace Anything [3], V-DPM [2]. Our Uni4R effectively captures the geometri…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

68 extracted references · 30 canonical work pages

  1. [1]

    St4rtrack: Simultaneous 4d reconstruction and tracking in the world

    Haiwen Feng, Junyi Zhang, Qianqian Wang, Yufei Ye, Pengcheng Yu, Michael J Black, Trevor Darrell, and Angjoo Kanazawa. St4rtrack: Simultaneous 4d reconstruction and tracking in the world. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 8503–8513, 2025

    work page 2025

  2. [2]

    V-dpm: 4d video reconstruction with dynamic point maps

    Edgar Sucar, Eldar Insafutdinov, Zihang Lai, and Andrea Vedaldi. V-dpm: 4d video reconstruction with dynamic point maps. arXiv preprint arXiv:2601.09499, 2026

    arXiv 2026

  3. [3]

    Trace anything: Representing any video in 4d via trajectory fields.arXiv preprint arXiv:2510.13802, 2025

    Xinhang Liu, Yuxi Xiao, Donny Y Chen, Jiashi Feng, Yu-Wing Tai, Chi-Keung Tang, and Bingyi Kang. Trace anything: Representing any video in 4d via trajectory fields.arXiv preprint arXiv:2510.13802, 2025

    arXiv 2025

  4. [4]

    Neural ordinary differential equations.Advances in neural information processing systems, 31, 2018

    Ricky TQ Chen, Yulia Rubanova, Jesse Bettencourt, and David K Duvenaud. Neural ordinary differential equations.Advances in neural information processing systems, 31, 2018. 9 APREPRINT- AUGUST11, 2026

    work page 2018

  5. [5]

    Augmented neural odes.Advances in neural information processing systems, 32, 2019

    Emilien Dupont, Arnaud Doucet, and Yee Whye Teh. Augmented neural odes.Advances in neural information processing systems, 32, 2019

    work page 2019

  6. [6]

    On second order behaviour in augmented neural odes.Advances in neural information processing systems, 33:5911–5921, 2020

    Alexander Norcliffe, Cristian Bodnar, Ben Day, Nikola Simidjievski, and Pietro Liò. On second order behaviour in augmented neural odes.Advances in neural information processing systems, 33:5911–5921, 2020

    2020

  7. [7]

    Flow matching for generative modeling

    Yaron Lipman, Ricky TQ Chen, Heli Ben-Hamu, Maximilian Nickel, and Matt Le. Flow matching for generative modeling. arXiv preprint arXiv:2210.02747, 2022

    arXiv 2022

  8. [8]

    Pix2vox: Context-aware 3d reconstruction from single and multi-view images

    Haozhe Xie, Hongxun Yao, Xiaoshuai Sun, Shangchen Zhou, and Shengping Zhang. Pix2vox: Context-aware 3d reconstruction from single and multi-view images. In2019 IEEE/CVF International Conference on Computer Vision (ICCV), pages 2690–2698. IEEE, 2019

    work page 2019

Show all 68 references
  1. [9]

    Garnet: Global-aware multi-view 3d reconstruction network and the cost-performance tradeoff.Pattern Recognition, 142:109674, 2023

    Zhenwei Zhu, Liying Yang, Xuxin Lin, Lin Yang, and Yanyan Liang. Garnet: Global-aware multi-view 3d reconstruction network and the cost-performance tradeoff.Pattern Recognition, 142:109674, 2023

    2023

  2. [10]

    Long-range grouping transformer for multi-view 3d reconstruction

    Liying Yang, Zhenwei Zhu, Xuxin Lin, Jian Nong, and Yanyan Liang. Long-range grouping transformer for multi-view 3d reconstruction. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 18257–18267, 2023

    2023

  3. [11]

    Umiformer: Mining the correlations between similar tokens for multi-view 3d reconstruction

    Zhenwei Zhu, Liying Yang, Ning Li, Chaohao Jiang, and Yanyan Liang. Umiformer: Mining the correlations between similar tokens for multi-view 3d reconstruction. In2023 IEEE/CVF International Conference on Computer Vision (ICCV), pages 18180–18189. IEEE, 2023

    2023

  4. [12]

    Wonder3d: Single image to 3d using cross-domain diffusion

    Xiaoxiao Long, Yuan-Chen Guo, Cheng Lin, Yuan Liu, Zhiyang Dou, Lingjie Liu, Yuexin Ma, Song-Hai Zhang, Marc Habermann, Christian Theobalt, et al. Wonder3d: Single image to 3d using cross-domain diffusion. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern...

    2024

  5. [13]

    Hunyuan3d 2.0: Scaling diffusion models for high resolution textured 3d assets generation.arXiv preprint arXiv:2501.12202, 2025

    Zibo Zhao, Zeqiang Lai, Qingxiang Lin, Yunfei Zhao, Haolin Liu, Shuhui Yang, Yifei Feng, Mingxin Yang, Sheng Zhang, Xianghui Yang, et al. Hunyuan3d 2.0: Scaling diffusion models for high resolution textured 3d assets generation.arXiv preprint arXiv:2501.12202, 2025

    2025 arXiv

  6. [14]

    Hunyuan3d-omni: A unified framework for controllable generation of 3d assets.arXiv preprint arXiv:2509.21245, 2025

    Team Hunyuan3D, Bowen Zhang, Chunchao Guo, Haolin Liu, Hongyu Yan, Huiwen Shi, Jingwei Huang, Junlin Yu, Kunhong Li, Penghao Wang, et al. Hunyuan3d-omni: A unified framework for controllable generation of 3d assets.arXiv preprint arXiv:2509.21245, 2025

    2025

  7. [15]

    Pc 2: Projection-conditioned point cloud diffusion for single-image 3d reconstruction

    Luke Melas-Kyriazi, Christian Rupprecht, and Andrea Vedaldi. Pc 2: Projection-conditioned point cloud diffusion for single-image 3d reconstruction. In2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 12923–12932. IEEE, 2023

    2023

  8. [16]

    Pointmc: Multi-view consistent encoding and center-global feature fusion for point clouds understanding

    Xinxing Yu, Ajian Liu, Sunyuan Qiang, Yuzhong Wang, Hui Ma, and Yanyan Liang. Pointmc: Multi-view consistent encoding and center-global feature fusion for point clouds understanding. InProceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 12169–12177, 2026

    2026

  9. [17]

    Facnet: Feature alignment fast point cloud completion network.Computational Visual Media, 11(1):141–157, 2025

    Xinxing Yu, Jianyi Li, Chi-Chong Wong, Chi-Man V ong, and Yanyan Liang. Facnet: Feature alignment fast point cloud completion network.Computational Visual Media, 11(1):141–157, 2025

    2025

  10. [18]

    Pointcsp: Cross-sample semantic propagation and stability preservation in self-supervised point cloud learning

    Xinxing Yu, Ajian Liu, Sunyuan Qiang, Hui Ma, Liying Yang, Yuzhong Wang, Zhi Rao, and Yanyan Liang. Pointcsp: Cross-sample semantic propagation and stability preservation in self-supervised point cloud learning. InProceedings of the IEEE/CVF Conference on Computer Vision and P...

    2026

  11. [19]

    Pointchr: Point cloud analysis via curvature-aware hyperbolic rectification

    Xinxing Yu, Liying Yang, Hao Mo, Hui Ma, Fang Kai, Ajian Liu, and Yanyan Liang. Pointchr: Point cloud analysis via curvature-aware hyperbolic rectification. InForty-third International Conference on Machine Learning, 2026

    2026

  12. [20]

    3d gaussian splatting for real-time radiance field rendering.ACM Trans

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler, George Drettakis, et al. 3d gaussian splatting for real-time radiance field rendering.ACM Trans. Graph., 42(4):139–1, 2023

    2023

  13. [21]

    Gs-lrm: Large reconstruction model for 3d gaussian splatting

    Kai Zhang, Sai Bi, Hao Tan, Yuanbo Xiangli, Nanxuan Zhao, Kalyan Sunkavalli, and Zexiang Xu. Gs-lrm: Large reconstruction model for 3d gaussian splatting. InEuropean Conference on Computer Vision, pages 1–19. Springer, 2024

    2024

  14. [22]

    Lgm: Large multi-view gaussian model for high-resolution 3d content creation

    Jiaxiang Tang, Zhaoxi Chen, Xiaokang Chen, Tengfei Wang, Gang Zeng, and Ziwei Liu. Lgm: Large multi-view gaussian model for high-resolution 3d content creation. InEuropean Conference on Computer Vision, pages 1–18. Springer, 2024

    2024

  15. [23]

    A unified approach for text-and image-guided 4d scene generation

    Yufeng Zheng, Xueting Li, Koki Nagano, Sifei Liu, Otmar Hilliges, and Shalini De Mello. A unified approach for text-and image-guided 4d scene generation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 7300–7309, 2024

    2024

  16. [24]

    Animate124: Animating one image to 4d dynamic scene.arXiv preprint arXiv:2311.14603, 2023

    Yuyang Zhao, Zhiwen Yan, Enze Xie, Lanqing Hong, Zhenguo Li, and Gim Hee Lee. Animate124: Animating one image to 4d dynamic scene.arXiv preprint arXiv:2311.14603, 2023

    2023 arXiv

  17. [25]

    Stag4d: Spatial-temporal anchored generative 4d gaussians

    Yifei Zeng, Yanqin Jiang, Siyu Zhu, Yuanxun Lu, Youtian Lin, Hao Zhu, Weiming Hu, Xun Cao, and Yao Yao. Stag4d: Spatial-temporal anchored generative 4d gaussians. InEuropean Conference on Computer Vision, pages 163–179. Springer, 2024

    2024

  18. [26]

    Sc4d: Sparse-controlled video-to-4d generation and motion transfer

    Zijie Wu, Chaohui Yu, Yanqin Jiang, Chenjie Cao, Fan Wang, and Xiang Bai. Sc4d: Sparse-controlled video-to-4d generation and motion transfer. InEuropean Conference on Computer Vision, pages 361–379. Springer, 2024

    2024

  19. [27]

    Not all frame features are equal: Video-to-4d generation via decoupling dynamic-static features

    Liying Yang, Chen Liu, Zhenwei Zhu, Ajian Liu, Hui Ma, Jian Nong, and Yanyan Liang. Not all frame features are equal: Video-to-4d generation via decoupling dynamic-static features. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 7494–7504, 2025...

    2025

  20. [28]

    Make-a-video: Text-to-video generation without text-video data.arXiv preprint arXiv:2209.14792, 2022

    Uriel Singer, Adam Polyak, Thomas Hayes, Xi Yin, Jie An, Songyang Zhang, Qiyuan Hu, Harry Yang, Oron Ashual, Oran Gafni, et al. Make-a-video: Text-to-video generation without text-video data.arXiv preprint arXiv:2209.14792, 2022

    2022 arXiv

  21. [29]

    Wan: Open and advanced large-scale video generative models.arXiv preprint arXiv:2503.20314, 2025

    Team Wan, Ang Wang, Baole Ai, Bin Wen, Chaojie Mao, Chen-Wei Xie, Di Chen, Feiwu Yu, Haiming Zhao, Jianxiao Yang, et al. Wan: Open and advanced large-scale video generative models.arXiv preprint arXiv:2503.20314, 2025

    2025 arXiv

  22. [30]

    Zero-1-to-3: Zero-shot one image to 3d object

    Ruoshi Liu, Rundi Wu, Basile Van Hoorick, Pavel Tokmakov, Sergey Zakharov, and Carl V ondrick. Zero-1-to-3: Zero-shot one image to 3d object. InProceedings of the IEEE/CVF international conference on computer vision, pages 9298–9309, 2023

    2023

  23. [31]

    Zero123++: a single image to consistent multi-view diffusion base model.arXiv preprint arXiv:2310.15110, 2023

    Ruoxi Shi, Hansheng Chen, Zhuoyang Zhang, Minghua Liu, Chao Xu, Xinyue Wei, Linghao Chen, Chong Zeng, and Hao Su. Zero123++: a single image to consistent multi-view diffusion base model.arXiv preprint arXiv:2310.15110, 2023

    2023 arXiv

  24. [32]

    Dreamfusion: Text-to-3d using 2d diffusion.arXiv preprint arXiv:2209.14988, 2022

    Ben Poole, Ajay Jain, Jonathan T Barron, and Ben Mildenhall. Dreamfusion: Text-to-3d using 2d diffusion.arXiv preprint arXiv:2209.14988, 2022

    2022 arXiv

  25. [33]

    Choreographing a world of dynamic objects.arXiv preprint arXiv:2601.04194, 2026

    Yanzhe Lyu, Chen Geng, Karthik Dharmarajan, Yunzhi Zhang, Hadi Alzayer, Shangzhe Wu, and Jiajun Wu. Choreographing a world of dynamic objects.arXiv preprint arXiv:2601.04194, 2026

    2026

  26. [34]

    L4gm: Large 4d gaussian reconstruction model.Advances in Neural Information Processing Systems, 37:56828–56858, 2024

    Jiawei Ren, Kevin Xie, Ashkan Mirzaei, Hanxue Liang, Xiaohui Zeng, Karsten Kreis, Ziwei Liu, Antonio Torralba, Sanja Fidler, Seung W Kim, et al. L4gm: Large 4d gaussian reconstruction model.Advances in Neural Information Processing Systems, 37:56828–56858, 2024

    2024

  27. [35]

    Seeu: Seeing the unseen world via 4d dynamics-aware generation.arXiv preprint arXiv:2512.03350, 2025

    Yu Yuan, Tharindu Wickremasinghe, Zeeshan Nadir, Xijun Wang, Yiheng Chi, and Stanley H Chan. Seeu: Seeing the unseen world via 4d dynamics-aware generation.arXiv preprint arXiv:2512.03350, 2025

    2025

  28. [36]

    Sv4d 2.0: Enhancing spatio-temporal consistency in multi-view video diffusion for high-quality 4d generation

    Chun-Han Yao, Yiming Xie, Vikram V oleti, Huaizu Jiang, and Varun Jampani. Sv4d 2.0: Enhancing spatio-temporal consistency in multi-view video diffusion for high-quality 4d generation. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 13248–13258, 2025

    2025

  29. [37]

    Spatial-temporal state propagation autoregressive model for 4d object generation.arXiv preprint arXiv:2602.18830, 2026

    Liying Yang, Jialun Liu, Jiakui Hu, Chenhao Guan, Haibin Huang, Fangqiu Yi, Chi Zhang, and Yanyan Liang. Spatial-temporal state propagation autoregressive model for 4d object generation.arXiv preprint arXiv:2602.18830, 2026

    2026

  30. [38]

    Dust3r: Geometric 3d vision made easy

    Shuzhe Wang, Vincent Leroy, Yohann Cabon, Boris Chidlovskii, and Jerome Revaud. Dust3r: Geometric 3d vision made easy. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 20697–20709, 2024

    2024

  31. [39]

    Fast3r: Towards 3d reconstruction of 1000+ images in one forward pass

    Jianing Yang, Alexander Sax, Kevin J Liang, Mikael Henaff, Hao Tang, Ang Cao, Joyce Chai, Franziska Meier, and Matt Feiszli. Fast3r: Towards 3d reconstruction of 1000+ images in one forward pass. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 21...

    2025

  32. [40]

    Vggt: Visual geometry grounded transformer

    Jianyuan Wang, Minghao Chen, Nikita Karaev, Andrea Vedaldi, Christian Rupprecht, and David Novotny. Vggt: Visual geometry grounded transformer. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 5294–5306, 2025

    2025

  33. [41]

    Yifan Wang, Jianjun Zhou, Haoyi Zhu, Wenzheng Chang, Yang Zhou, Zizun Li, Junyi Chen, Jiangmiao Pang, Chunhua Shen, and Tong He.π 3: Permutation-equivariant visual geometry learning.arXiv preprint arXiv:2507.13347, 2025

    2025 arXiv

  34. [42]

    Dynamic point maps: A versatile representation for dynamic 3d reconstruction

    Edgar Sucar, Zihang Lai, Eldar Insafutdinov, and Andrea Vedaldi. Dynamic point maps: A versatile representation for dynamic 3d reconstruction. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 7295–7305, 2025

    2025

  35. [43]

    Mcgs: Markov chain gaussian splatting for dynamic scenes reconstruction

    Yuzhong Wang, Wenmin Wang, Shixiong Zhang, Xinxing Yu, and Zhongheng Chen. Mcgs: Markov chain gaussian splatting for dynamic scenes reconstruction. InProceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 10341–10348, 2026

    2026

  36. [44]

    Cotracker: It is better to track together

    Nikita Karaev, Ignacio Rocco, Benjamin Graham, Natalia Neverova, Andrea Vedaldi, and Christian Rupprecht. Cotracker: It is better to track together. InEuropean conference on computer vision, pages 18–35. Springer, 2024

    2024

  37. [45]

    Cotracker3: Simpler and better point tracking by pseudo-labelling real videos

    Nikita Karaev, Yuri Makarov, Jianyuan Wang, Natalia Neverova, Andrea Vedaldi, and Christian Rupprecht. Cotracker3: Simpler and better point tracking by pseudo-labelling real videos. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 6013–6022, 2025

    2025

  38. [46]

    Spatialtracker: Tracking any 2d pixels in 3d space

    Yuxi Xiao, Qianqian Wang, Shangzhan Zhang, Nan Xue, Sida Peng, Yujun Shen, and Xiaowei Zhou. Spatialtracker: Tracking any 2d pixels in 3d space. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20406–20417, 2024

    2024

  39. [47]

    Spatialtrackerv2: Advancing 3d point tracking with explicit camera motion

    Yuxi Xiao, Jianyuan Wang, Nan Xue, Nikita Karaev, Yuri Makarov, Bingyi Kang, Xing Zhu, Hujun Bao, Yujun Shen, and Xiaowei Zhou. Spatialtrackerv2: Advancing 3d point tracking with explicit camera motion. InProceedings of the IEEE/CVF International Conference on Computer Vision,...

    2025

  40. [48]

    Pomato: Marrying pointmap matching with temporal motions for dynamic 3d reconstruction

    Songyan Zhang, Yongtao Ge, Jinyuan Tian, Guangkai Xu, Hao Chen, Chen Lv, and Chunhua Shen. Pomato: Marrying pointmap matching with temporal motions for dynamic 3d reconstruction. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 5680–5689, 2025

    2025

  41. [49]

    Flow4r: Unifying 4d reconstruction and tracking with scene flow, 2026

    Shenhan Qian, Ganlin Zhang, Shangzhe Wu, and Daniel Cremers. Flow4r: Unifying 4d reconstruction and tracking with scene flow, 2026

    2026

  42. [50]

    Efficiently reconstructing dynamic scenes one d4rt at a time.arXiv preprint arXiv:2512.08924, 2025

    Chuhan Zhang, Guillaume Le Moing, Skanda Koppula, Ignacio Rocco, Liliane Momeni, Junyu Xie, Shuyang Sun, Rahul Sukthankar, Joëlle K Barral, Raia Hadsell, et al. Efficiently reconstructing dynamic scenes one d4rt at a time.arXiv preprint arXiv:2512.08924, 2025. 11 APREPRINT- AU...

    2025

  43. [51]

    Continuous-time flows for efficient inference and density estimation

    Changyou Chen, Chunyuan Li, Liqun Chen, Wenlin Wang, Yunchen Pu, and Lawrence Carin Duke. Continuous-time flows for efficient inference and density estimation. InInternational Conference on Machine Learning, pages 824–833. PMLR, 2018

    2018

  44. [52]

    Ffjord: Free-form continuous dynamics for scalable reversible generative models.arXiv preprint arXiv:1810.01367, 2018

    Will Grathwohl, Ricky TQ Chen, Jesse Bettencourt, Ilya Sutskever, and David Duvenaud. Ffjord: Free-form continuous dynamics for scalable reversible generative models.arXiv preprint arXiv:1810.01367, 2018

    2018 arXiv

  45. [53]

    Improving and generalizing flow-based generative models with minibatch optimal transport.arXiv preprint arXiv:2302.00482, 2023

    Alexander Tong, Kilian Fatras, Nikolay Malkin, Guillaume Huguet, Yanlei Zhang, Jarrid Rector-Brooks, Guy Wolf, and Yoshua Bengio. Improving and generalizing flow-based generative models with minibatch optimal transport.arXiv preprint arXiv:2302.00482, 2023

    2023 arXiv

  46. [54]

    A computational fluid mechanics solution to the monge-kantorovich mass transfer problem.Numerische Mathematik, 84(3):375–393, 2000

    Jean-David Benamou and Yann Brenier. A computational fluid mechanics solution to the monge-kantorovich mass transfer problem.Numerische Mathematik, 84(3):375–393, 2000

    2000

  47. [55]

    Flow straight and fast: Learning to generate and transfer data with rectified flow.arXiv preprint arXiv:2209.03003, 2022

    Xingchao Liu, Chengyue Gong, and Qiang Liu. Flow straight and fast: Learning to generate and transfer data with rectified flow.arXiv preprint arXiv:2209.03003, 2022

    2022 arXiv

  48. [56]

    Flashattention: Fast and memory-efficient exact attention with io-awareness.Advances in neural information processing systems, 35:16344–16359, 2022

    Tri Dao, Dan Fu, Stefano Ermon, Atri Rudra, and Christopher Ré. Flashattention: Fast and memory-efficient exact attention with io-awareness.Advances in neural information processing systems, 35:16344–16359, 2022

    2022

  49. [57]

    Morgan Kaufmann, 2002

    Gerald E Farin.Curves and surfaces for CAGD: a practical guide. Morgan Kaufmann, 2002

    2002

  50. [58]

    Kubric: A scalable dataset generator

    Klaus Greff, Francois Belletti, Lucas Beyer, Carl Doersch, Yilun Du, Daniel Duckworth, David J Fleet, Dan Gnanapragasam, Florian Golemo, Charles Herrmann, et al. Kubric: A scalable dataset generator. InProceedings of the IEEE/CVF conference on computer vision and pattern recog...

    2022

  51. [59]

    The coordination of arm movements: an experimentally confirmed mathematical model

    Tamar Flash and Neville Hogan. The coordination of arm movements: an experimentally confirmed mathematical model. Journal of neuroscience, 5(7):1688–1703, 1985

    1985

  52. [60]

    As-rigid-as-possible surface modeling

    Olga Sorkine, Marc Alexa, et al. As-rigid-as-possible surface modeling. InSymposium on Geometry processing, volume 4, pages 109–116, 2007

    2007

  53. [61]

    Pointodyssey: A large-scale synthetic dataset for long-term point tracking

    Yang Zheng, Adam W Harley, Bokui Shen, Gordon Wetzstein, and Leonidas J Guibas. Pointodyssey: A large-scale synthetic dataset for long-term point tracking. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 19855–19865, 2023

    2023

  54. [62]

    Dynamicstereo: Consistent dynamic depth from stereo videos

    Nikita Karaev, Ignacio Rocco, Benjamin Graham, Natalia Neverova, Andrea Vedaldi, and Christian Rupprecht. Dynamicstereo: Consistent dynamic depth from stereo videos. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 13229–13239, 2023

    2023

  55. [63]

    Aria digital twin: A new benchmark dataset for egocentric 3d machine perception

    Xiaqing Pan, Nicholas Charron, Yongqian Yang, Scott Peters, Thomas Whelan, Chen Kong, Omkar Parkhi, Richard Newcombe, and Yuheng Carl Ren. Aria digital twin: A new benchmark dataset for egocentric 3d machine perception. InProceedings of the IEEE/CVF International Conference on...

    2023

  56. [64]

    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. In2015 IEEE International Conference on Computer Vision (ICCV), pages 3334–3342, 2015

    2015

  57. [65]

    A benchmark dataset and evaluation methodology for video object segmentation

    Federico Perazzi, Jordi Pont-Tuset, Brian McWilliams, Luc Van Gool, Markus Gross, and Alexander Sorkine-Hornung. A benchmark dataset and evaluation methodology for video object segmentation. InProceedings of the IEEE conference on computer vision and pattern recognition, pages...

    2016

  58. [66]

    Tapvid-3d: A benchmark for tracking any point in 3d.Advances in Neural Information Processing Systems, 37:82149–82165, 2024

    Skanda Koppula, Ignacio Rocco, Yi Yang, Joe Heyward, Joao Carreira, Andrew Zisserman, Gabriel Brostow, and Carl Doersch. Tapvid-3d: A benchmark for tracking any point in 3d.Advances in Neural Information Processing Systems, 37:82149–82165, 2024

    2024

  59. [67]

    Monst3r: A simple approach for estimating geometry in the presence of motion.arXiv preprint arXiv:2410.03825, 2024

    Junyi Zhang, Charles Herrmann, Junhwa Hur, Varun Jampani, Trevor Darrell, Forrester Cole, Deqing Sun, and Ming-Hsuan Yang. Monst3r: A simple approach for estimating geometry in the presence of motion.arXiv preprint arXiv:2410.03825, 2024

    arXiv

  60. [68]

    Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography.Communications of the ACM, 24(6):381–395, 1981

    Martin A Fischler and Robert C Bolles. Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography.Communications of the ACM, 24(6):381–395, 1981. 12

    1981

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.