Pith. sign in

REVIEW 3 major objections 5 minor 116 references

Gaussians on Fire: High-Frequency Reconstruction of Flames

T0 review · 3 major / 5 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read A pipeline reconstructs high-frequency 3D flame dynamics from just three calibrated, hardware-synchronized cameras, using Gaussian splats that each carry a lifetime and a linear velocity.

desk verdict Valuable dataset and capture rig; the central 3D claim needs a held-out-view test before it is proven. read the letter →

arxiv 2511.22459 v2 pith:45FRVK6L submitted 2025-11-27 cs.CV

Jakob Nazarenus , Dominik Michels , Wojtek Palubicki , Simin Kou , Fang-Lue Zhang , Sören Pirk , Reinhard Koch This is my paper · ORCID
classification cs.CV
keywords dynamicscenereconstruction3DGaussiansplattingfireopticalflowfusionsparsemulti-viewvideotemporalparameterizationrollingshuttersynchronizationnovelviewsynthesis
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 high-frequency flame dynamics can be reconstructed as photorealistic 4D models from only three calibrated, hardware-synchronized consumer cameras, an under-constrained setting where previous dynamic Gaussian methods collapse into per-camera 2D reconstructions. Its strategy is to split the problem: reconstruct the static background with depth-regularized 3D Gaussian splats, then model the fire as a cloud of Gaussians whose positions are seeded by a 3D flow field fused from per-view 2D optical flow, and whose temporal behavior is explicitly parameterized by a lifetime and a linear velocity per splat. On 17 real fire sequences, the method reports fire-region PSNR of 27.46 and SSIM of 0.847, with a depth error of 0.0412, against 18.24 and 25.51 PSNR for two existing dynamic-Gaussian baselines. If correct, it would make high-fidelity fire capture practical with cheap hardware and would expose the flame motion field directly for analysis rather than hiding it inside a learned scene representation.

What carries the argument

The load-bearing machinery has three coupled parts. First, the static/dynamic split: the background is reconstructed once with vanilla 3D Gaussian splats whose geometry is initialized by dense stereo depth and regularized by aligned monocular depth, so the under-constrained three-view problem never touches the flames. Second, the flow-fusion initialization: per-view dense optical flow is back-projected into a voxel grid, and at each voxel the unknown 3D velocity F must satisfy (F - u_i) . u_i = 0 for each camera's back-projected 2D flow u_i; because three constraints do not determine a 3D vector, a Tikhonov term with a voxel-dependent weight selects a velocity, and the resulting volume seeds

What would settle it

Run the pipeline on a synthetic fire sequence with known ground-truth 3D velocity, using only three views; measure per-Gaussian velocity error against ground truth while keeping PSNR high. If the recovered trajectories diverge substantially from ground truth even though the images match, the central claim of reconstructing flame dynamics is falsified — the flow-initialized Gaussians would be fitting appearance, not motion.

Watch

Extended reading notes

Core claim

The discovery is that a dynamic 3D Gaussian representation can reconstruct fire from extremely sparse views, provided the Gaussians' temporal parameters are initialized from observed motion rather than learned from scratch. The key move is to treat each flame Gaussian as a particle: it has a birth time, a lifetime, and a constant velocity, so its world-space trajectory is x(t) = x0 + (t - t_mu)v. This parameterization is seeded by fusing dense 2D optical flow from all three cameras onto a voxel grid; at each voxel the back-projected 2D flows constrain the unknown 3D velocity to satisfy a projection constraint for each camera, and a regularization term selects the smallest admissible velocity

Load-bearing premise

The whole method stands or falls on the 3D flow field stitched together from three 2D motion estimates: with only three cameras, each voxel is under-constrained, and the paper's own limitation section (Sec. 6) concedes that optical-flow errors in bright or low-texture flame regions propagate directly into the 3D motion field and into every Gaussian seeded from it.

Editorial extensions

If this is right

  • Fire capture moves from dense camera arrays or physics simulators to three synchronized consumer cameras, lowering hardware cost and setup complexity.
  • The reconstructed model can be rendered from novel viewpoints outside the training rig and at arbitrary times, because each Gaussian's position and opacity are defined for all times by its velocity and lifetime.
  • Flame motion becomes an explicit, queryable quantity: every Gaussian's velocity is a number, not a learned latent, so trajectories can be exported for analysis, animation, or reactive systems.
  • Sparse-view dynamic Gaussian reconstruction is substantially more stable when velocities are initialized from observed optical flow rather than randomly, per the paper's RandFlow ablation.
  • Separating static background from dynamic fire during training yields not only sharper fire rendering but also geometrically plausible depth, whereas joint optimization collapses depth accuracy.

Reading between the lines

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

  • The same split-and-seed recipe should transfer to other semi-transparent volumetric phenomena such as smoke, steam, dust, or spray; the paper only evaluates combustion by-products, not whether the velocity-carrying Gaussian form is the general mechanism.
  • The explicit velocity field could be validated against combustion physics (buoyant rise profiles, turbulent spectra) as a way to separate reconstructed motion from photometric artifacts; the paper does not do this comparison.
  • A testable scaling prediction follows: if the flow fusion is the active ingredient, reconstruction quality in fire regions should degrade smoothly as cameras are removed from three to two, and the regularization bias toward small velocities should become visible in the recovered trajectories; the paper reports only the three-view case.
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

3 major / 5 minor

Summary. The paper proposes Gaussians on Fire, a pipeline for reconstructing 4D fire from three calibrated, hardware-synchronized consumer cameras. The static background is reconstructed with vanilla 3DGS initialized from stereo depth and regularized by aligned DepthAnythingV2 monocular depth; the dynamic fire is represented with FreeTimeGS-style Gaussians carrying linear velocity, lifetime, and temporal center, initialized from a Tikhonov-regularized fusion of per-view dense optical flow into a voxel grid. The authors introduce a real dataset of 17 fire scenes at 400 Hz, a custom ESP32/LED synchronization rig, and a rolling-shutter model. They report PSNR_flame 27.46 ± 1.76 versus 18.24/25.51 for the two 4DGS baselines, RMSE_depth 0.0412 versus ~0.27, and ablations for sync, rolling shutter, flow initialization, and static pre-training.

Significance. If the central claim were firmly established, this would be a useful contribution: it would show that high-frequency dynamic fire can be reconstructed from a minimal three-camera rig and that explicit per-Gaussian velocity and lifetime parameters are a practical representation for emissive, transparent scenes. Strengths include a real captured dataset, a concrete hardware synchronization design with measured temporal precision, a rolling-shutter correction model, and ablations that isolate the main components. The code and data are promised for release, which would help reproducibility. However, the evidence as presented does not yet establish genuine 3D generalization across viewpoints, and the only quantitative geometric metric is partially circular.

major comments (3)
  1. [Sec. 5, Tables 1–2, Supp. Sec. 8] All quantitative metrics are computed on temporally held-out frames rendered from the same three training viewpoints. The supplementary acknowledges that 4DGS (Wu et al.) degenerates to a separate per-camera 2D video representation, and PSNR_flame/SSIM_flame would still reward such a 2D model on a temporal test. The central claim is '3D reconstruction from three views', so the evaluation must include a held-out viewpoint: for example, train on two views and test on the third, or render a calibrated novel camera path and compare against a held-out view quantitatively. The qualitative novel-view video is a useful start but does not separate true view interpolation from appearance memory.
  2. [Sec. 3.1 Eq. (3), Sec. 5 Table 1] RMSE_depth is computed against DepthAnythingV2 predictions after linear alignment, while the static reconstruction is trained with λ_depth increased by a factor of 100 against those same aligned predictions. The large gap between 0.0412 and ~0.27 therefore partly measures agreement with the training regularizer, not geometric accuracy. For the transparent/emissive fire region, monocular depth is an unreliable geometric reference. Please report depth against a geometrically grounded reference not used in training, or at least evaluate with the depth regularization switched off and clearly state the change in the comparison.
  3. [Sec. 3.2 Eqs. (5)–(8), Table 2 (RandFlow)] The fused 3D flow is underdetermined: with m=3 cameras, each voxel provides only three scalar projection constraints on a 3D vector, and closure relies on Tikhonov regularization with hand-set α0. The RandFlow ablation retains PSNR_flame 27.01 versus 27.46, showing that the photometric loss largely compensates for incorrect velocity initialization. Since the paper claims reconstruction of flame dynamics, please provide a direct evaluation of the velocity field—e.g., predicted 3D trajectories projected into a held-out view and compared with tracked flame features, or comparison with a physically motivated vertical-motion prior. The paper's own limitation statement in Sec. 6 concedes that optical-flow errors propagate into the 3D motion field, so an independent validation of the flow component is needed.
minor comments (5)
  1. [Sec. 5] Typographical issues: 'RSMEdepth' should be 'RMSE_depth', 'archives' should be 'achieves', and 'high offerfitting' should be 'high overfitting'.
  2. [Eq. (5)] The notation π^{-1}_i(..., x) is nonstandard, and the expression f_i(π(x)) + π_i(x) mixes 2D and 3D operands. Please rewrite with explicit coordinates or provide a reference for this back-projection operation.
  3. [Secs. 3.3–3.4] The measured readout time is stated as 2.85 μs per line in Sec. 3.3 and t_row ≈ 3 μs in Sec. 3.4; clarify whether these are the same quantity or different measurements.
  4. [References] References [91] and [92] appear to be the same paper (Wu et al., CVPR 2024) listed twice. Please deduplicate.
  5. [Supp. Sec. 10] The description 'increased the depth loss by a factor of 100, yielding an exponentially decreasing contribution with an initial value of 100 and a final value of 1' is ambiguous about the scheduling. Specify the exact annealing schedule.

Circularity Check

1 steps flagged · score 5.0 of 10

Depth evaluation is circular: RMSE_depth compares against the same monocular-depth predictor used as the static-scene regularizer; photometric reconstruction itself is tested on held-out frames.

  1. fitted input called prediction [Sec. 3.1 (Eq. 3) and Sec. 5 (Tab. 1, Fig. 9)]
    "During training, the aligned monocular depths D′mono are incorporated as a regularization term ... Ldepth = λdepth/wh Σ ||D′mono(i,j)−Drender(i,j)||1 ... We increase the default regularization parameter value by a factor of 100. ... To evaluate the plausibility of the rendered depth, we compare it against a predicted monocular depth after linear alignment and report this score as RSMEdepth."

    The reported RMSE_depth metric uses DepthAnythingV2 monocular depth (after linear alignment) as the reference. That is the same signal used as the supervision target in Eq. 3, with λ_depth raised by a factor of 100 during static-scene training. A low RMSE_depth therefore largely indicates agreement with the training regularizer, not agreement with independent 3D geometry. For the dynamic fire region, monocular depth is not a reliable geometric reference, and for the static region the low error is partly by construction. Thus the paper's quantitative evidence for 'geometrically plausible depth' and multi-view 3D consistency is partly circular.

full rationale

The photometric reconstruction claim is not circular: PSNR_flame/SSIM_flame are computed on temporally held-out frames (every 8th frame), which is a legitimate test of temporal generalization, and the flow-initialization is ablated (RandFlow) rather than assumed. The paper also does not rely on a load-bearing self-citation chain: MEMFOF, DepthAnythingV2, and FreeTimeGS are external methods, not author-defined constraints. However, the only quantitative geometry metric, RMSE_depth, is evaluated against the same monocular-depth predictor used to regularize the static scene, making that specific evaluation partially self-referential. The paper's supplementary even notes that one baseline degenerates to a per-camera 2D video, but no quantitative experiment separates the proposed method from a more graceful version of the same failure; this is a validity concern rather than a derivation-level circularity. Overall, the central contribution retains independent content, but the depth evidence is weakened by construction, yielding a moderate circularity score.

Assumptions & free parameters 6 free parameters · 6 assumptions · 1 invented entities

The method deliberately imports its two main priors - 2D optical flow (MEMFOF) and monocular depth (DepthAnythingV2) - from learned models, and its dynamic representation from FreeTimeGS. The genuinely new numbers introduced (flow-Tikhonov α0, x100 depth weight, 4-frame lifespan, 4 px/ms motion cap, voxel size s) are hand-chosen and mostly unreported in value; none is justified by a physical constraint. No new physical entity is postulated; the only invented artifact is the LED sync hardware, which carries observable evidence in every frame.

free parameters (6)
  • α0 (Tikhonov regularization strength, Eq. 7) = not reported
    Hand-set weight that closes the under-determined 3-view flow solve; its value is never stated although it directly shapes the velocity field initializing all dynamic Gaussians.
  • λ_depth scaling factor = 100 -> 1 (exponential decay)
    Depth-regularization weight raised 100x over the standard 3DGS value (Sec. 3.1) to prevent static-geometry collapse under three views; hand-chosen and scene-independent.
  • Gaussian lifespan initialization = 4 x frame time ~ 10 ms
    Chosen 'for smooth motion and to prevent overfitting to single frames' (Sec. 3.2); sets the temporal smoothing kernel of every dynamic splat and directly produces the acknowledged oversmoothing.
  • Maximum screen-space motion cap = 4 px/ms
    Hand-set cap used to justify dropping the denominator in the rolling-shutter delay approximation (∇t·v ≈ 11e-3, Sec. 3.3); device-specific and not derived from a physical constraint.
  • Voxel size s (flow-fusion grid) = not reported
    Discretization of the volumetric flow field in Sec. 3.2 controls Gaussian density and the random de-moireing offset Δx; the value is omitted from the paper.
  • Velocity spatial learning-rate reduction = x10^-2
    Locks the optimized velocity near its flow-initialized value (Sec. 11); this is why the claimed motion output largely reflects the input prior rather than an independently measured 3D motion.
assumptions (6)
  • domain assumption Minimum-intensity projection over time removes fire: flames are locally brighter than their backdrop and the background is static during the sequence.
    Sec. 3.1 / Fig. 3. If a flame region is darker than the background, or the background moves (smoke, wind-blown foliage), the derived 'static' frames and the COLMAP calibration are corrupted, breaking both reconstruction stages.
  • domain assumption The 3-view projection constraint (F-u_i)·u_i = 0 together with Tikhonov regularization recovers a usable 3D flow field (Eqs. 5-8).
    Sec. 3.2. The inversion is under-determined (3 scalar constraints per voxel for a 3-vector F) and neglects perspective and surface-orientation effects; the paper itself notes optical-flow errors propagate into the 3D motion field.
  • domain assumption FreeTimeGS linear-velocity plus Gaussian-lifetime representation is expressive enough for flame dynamics.
    Sec. 3.2; limitation admitted in Sec. 6 ('may not fully capture the complex, turbulent motion ... temporal blurring or oversmoothing'). This tensions with the 'high-frequency' claim.
  • domain assumption DepthAnythingV2 monocular depths, after per-camera affine alignment, are valid regularizers and valid references for evaluating rendered depth.
    Sec. 3.1 Eqs. 1-3 and Sec. 5. The same predictor is used for training (λ_depth = 100) and for the RMSE_depth metric, a circularity in the geometry evaluation.
  • domain assumption Rolling-shutter delay is dominated by line readout (2.85 μs/line) and is locally linear; the denominator 1 - ∇t·v ≈ 0.989 can be dropped.
    Sec. 3.3 Eq. 9. Device-specific approximation; the GlobalShutter ablation shows it changes PSNR_flame by only ~0.02 dB, so this modeling choice is empirically a minor correction.
  • domain assumption COLMAP camera poses estimated on fire-free static frames remain valid for the dynamic frames.
    Sec. 3.1. Requires rigid, fixed cameras; vibration or thermal drift over the 15 s captures is neglected.
invented entities (1)
  • ESP32 LED synchronization rig (Gray-code frame counter + 5 toggling COB LED strips) independent evidence
    purpose: Provides frame-identity and sub-frame (~15 μs) temporal alignment for three unsynchronized rolling-shutter consumer cameras at 400 fps.
    Not a postulated physical entity but a hardware invention with a falsifiable handle: the LED patterns are visible in every captured frame, timing errors are readable directly from the images, and the NoSync ablation quantifies the cost of losing this alignment. Independent verification would re-derive the ~15 μs precision from the observed brightness transitions.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Gaussians on Fire: High-Frequency Reconstruction of Flames." pith.science (2026). https://pith.science/paper/45FRVK6L

@misc{pith2026251122459,
  author       = {Pith},
  title        = {Pith review of: Gaussians on Fire: High-Frequency Reconstruction of Flames},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/45FRVK6L}},
  note         = {Machine review of arXiv:2511.22459}
}
read the original abstract

We propose a method to reconstruct dynamic fire in 3D from a limited set of camera views with a Gaussian-based spatiotemporal representation. Capturing and reconstructing fire and its dynamics is highly challenging due to its volatile nature, transparent quality, and multitude of high-frequency features. Despite these challenges, we aim to reconstruct fire from only three views, which consequently requires solving for under-constrained geometry. We solve this by separating the static background from the dynamic fire region by combining dense multi-view stereo images with monocular depth priors. The fire is initialized as a 3D flow field, obtained by fusing per-view dense optical flow projections. To capture the high frequency features of fire, each 3D Gaussian encodes a lifetime and linear velocity to match the dense optical flow. To ensure sub-frame temporal alignment across cameras we employ a custom hardware synchronization pattern -- allowing us to reconstruct fire with affordable commodity hardware. Our quantitative and qualitative validations across numerous reconstruction experiments demonstrate robust performance for diverse and challenging real fire scenarios.

Figures

Figures reproduced from arXiv: 2511.22459 by the authors.

Figure 1
Figure 1. Temporal progression of a flame encoded as 3D Gaussian splats. The sequence shows the evolution from an initial small [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of our pipeline: Given the video of a fire [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Fire removal for dynamic scenes: we stack all frames [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Intermediate data used for scene reconstruction: (a) In [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Multi-view capture of a fire: we capture the RGB im [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]
Figure 6
Figure 6. Figure 6: Illustration of rolling-shutter effects. (a) Image rendered [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]
Figure 7
Figure 7. Figure 7: Capture setup and synchronization hardware: Three [PITH_FULL_IMAGE:figures/full_fig_p005_7.png]
Figure 8
Figure 8. Figure 8: Rendered test views from the real-world fire dataset for different fuel types: Sequence (a) depicts a propane jet, (b) is comprised [PITH_FULL_IMAGE:figures/full_fig_p006_8.png]
Figure 9
Figure 9. Figure 9: Qualitative comparison of our method against 4DGS ( [PITH_FULL_IMAGE:figures/full_fig_p007_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

116 extracted references · 2 linked inside Pith

  1. [1]

    Per-gaussian embedding- based deformation for deformable 3d gaussian splatting

    Jeongmin Bae, Seoha Kim, Youngsik Yun, Hahyun Lee, Gun Bang, and Youngjung Uh. Per-gaussian embedding- based deformation for deformable 3d gaussian splatting

  2. [2]

    Narasimhan

    Aayush Bansal, Minh V o, Yaser Sheikh, Deva Ramanan, and Srinivasa G. Narasimhan. 4d visualization of dynamic events from unconstrained multi-view videos. InCVPR,

  3. [3]

    Memfof: High-resolution training for memory-efficient multi-frame optical flow estimation

    Vladislav Bargatin, Egor Chistov, Alexander Yakovenko, and Dmitriy Vatolin. Memfof: High-resolution training for memory-efficient multi-frame optical flow estimation. In ICCV, 2025. 2, 4

    2025

  4. [4]

    SURF: Speeded up robust features

    Herbert Bay, Tinne Tuytelaars, and Luc Van Gool. SURF: Speeded up robust features. InECCV, 2006. 2

    2006

  5. [5]

    Black and Padmanabhan Anandan

    Michael J. Black and Padmanabhan Anandan. A framework for the robust estimation of optical flow. InICCV, 1993. 2

    1993

  6. [6]

    Patchmatch stereo-stereo matching with slanted support windows

    Michael Bleyer, Christoph Rhemann, and Carsten Rother. Patchmatch stereo-stereo matching with slanted support windows. InBmvc, pages 1–11, 2011. 4

    2011

  7. [7]

    Depth pro: Sharp monocular metric depth in less than a second

    Alexey Bochkovskiy, Ama ¨el Delaunoy, Hugo Germain, Marcel Santos, Yichao Zhou, Stephan Richter, and Vladlen Koltun. Depth pro: Sharp monocular metric depth in less than a second. InICLR, pages 75602–75637, 2025. 2

    2025

  8. [8]

    Deepdeform: Learning non-rigid RGB- D reconstruction with semi-supervised data

    Aljaz Bozic, Michael Zollh ¨ofer, Christian Theobalt, and Matthias Nießner. Deepdeform: Learning non-rigid RGB- D reconstruction with semi-supervised data. InCVPR,

Show all 116 references
  1. [9]

    Recovering non-rigid 3d shape from image streams

    Christoph Bregler, Aaron Hertzmann, and Henning Bier- mann. Recovering non-rigid 3d shape from image streams. InCVPR, 2000. 3

    2000

  2. [10]

    High accuracy optical flow estimation based on a theory for warping

    Thomas Brox, Andr ´es Bruhn, Nils Papenberg, and Joachim Weickert. High accuracy optical flow estimation based on a theory for warping. InECCV, 2004. 2

    2004

  3. [11]

    Large displacement optical flow

    Thomas Brox, Christoph Bregler, and Jitendra Malik. Large displacement optical flow. InCVPR, 2009. 2

    2009

  4. [12]

    Immersive light field video with a layered mesh representation.ACM TOG, 39(4), 2020

    Michael Broxton, John Flynn, Ryan Overbeck, Daniel Er- ickson, Peter Hedman, Matthew Duvall, Jason Dourgarian, Jay Busch, Matt Whalen, and Paul Debevec. Immersive light field video with a layered mesh representation.ACM TOG, 39(4), 2020. 3

    2020

  5. [13]

    Hexplane: A fast representa- tion for dynamic scenes

    Ang Cao and Justin Johnson. Hexplane: A fast representa- tion for dynamic scenes. InCVPR, pages 130–141, 2023. 3

    2023

  6. [14]

    Black, Otmar Hilliges, and Andreas Geiger

    Xu Chen, Yufeng Zheng, Michael J. Black, Otmar Hilliges, and Andreas Geiger. SNARF: Differentiable forward skin- ning for animating non-rigid neural implicit shapes. In ICCV, 2021. 3

    2021

  7. [15]

    Physics informed neural fields for smoke reconstruction with sparse data.ACM TOG, 41(4):119:1–119:14, 2022

    Mengyu Chu, Lingjie Liu, Quan Zheng, Erik Franz, Hans- Peter Seidel, Christian Theobalt, and Rhaleb Zayer. Physics informed neural fields for smoke reconstruction with sparse data.ACM TOG, 41(4):119:1–119:14, 2022. 1, 2, 3

    2022

  8. [16]

    A simple prior- free method for non-rigid structure-from-motion factoriza- tion.IJCV, 107(2), 2014

    Yuchao Dai, Hongdong Li, and Mingyi He. A simple prior- free method for non-rigid structure-from-motion factoriza- tion.IJCV, 107(2), 2014. 3

    2014

  9. [17]

    Superpoint: Self-supervised interest point detection and description

    Daniel DeTone, Tomasz Malisiewicz, and Andrew Rabi- novich. Superpoint: Self-supervised interest point detection and description. InCVPRW, 2018. 2

    2018

  10. [18]

    Tap-vid: A benchmark for track- ing any point in a video

    Carl Doersch, Ankush Gupta, Larisa Markeeva, Adri `a Re- casens, Lucas Smaira, Yusuf Aytar, Jo˜ao Carreira, Andrew Zisserman, and Yi Yang. Tap-vid: A benchmark for track- ing any point in a video. InNeurIPS, 2022. 2

    2022

  11. [19]

    Tapir: Tracking any point with per-frame initialization and temporal refinement

    Carl Doersch, Yi Yang, Mel Vecerik, Dilara Gokay, Ankush Gupta, Yusuf Aytar, Jo˜ao Carreira, and Andrew Zisserman. Tapir: Tracking any point with per-frame initialization and temporal refinement. InICCV, 2023

    2023

  12. [20]

    Bootstap: Bootstrapped training for tracking-any-point

    Carl Doersch, Pauline Luc, Yi Yang, Dilara Gokay, Skanda Koppula, Ankush Gupta, Joseph Heyward, Ignacio Rocco, Ross Goroshin, Joao Carreira, et al. Bootstap: Bootstrapped training for tracking-any-point. InACCV, 2024. 2

    2024

  13. [21]

    Degtyarev, Philip L

    Mingsong Dou, Sameh Khamis, Yury G. Degtyarev, Philip L. Davidson, Sean Fanello, Adarsh Kowdle, Ser- gio Orts, Christoph Rhemann, David Kim, Jonathan Taylor, Pushmeet Kohli, Vladimir Tankovich, and Shahram Izadi. Fusion4d: Real-time volumetric capture of dynamic scenes. ACM TO...

    2016

  14. [22]

    Tenen- baum, and Jiajun Wu

    Yilun Du, Yinan Zhang, Hong-Xing Yu, Joshua B. Tenen- baum, and Jiajun Wu. Neural radiance flow for 4d view synthesis and video processing. InICCV, 2021. 3

    2021

  15. [23]

    4d-rotor gaussian splat- ting: Towards efficient novel view synthesis for dynamic scenes

    Yuanxing Duan, Fangyin Wei, Qiyu Dai, Yuhang He, Wen- zheng Chen, and Baoquan Chen. 4d-rotor gaussian splat- ting: Towards efficient novel view synthesis for dynamic scenes. InSIGGRAPH 2024 Conference Papers, 2024. 3

    2024

  16. [24]

    MASt3r-sfm: a fully-integrated solution for unconstrained structure-from-motion

    Bardienus Pieter Duisterhof, Lojze Zust, Philippe Weinza- epfel, Vincent Leroy, Yohann Cabon, and Jerome Revaud. MASt3r-sfm: a fully-integrated solution for unconstrained structure-from-motion. InInternational Conference on 3D Vision 2025, 2025. 2, 3

    2025

  17. [25]

    Black, Trevor Darrell, and Angjoo Kanazawa

    Haiwen Feng*, Junyi Zhang*, Qianqian Wang, Yufei Ye, Pengcheng Yu, Michael J. Black, Trevor Darrell, and Angjoo Kanazawa. St4rtrack: Simultaneous 4d reconstruc- tion and tracking in the world. InICCV, 2025. 2, 3

    2025

  18. [26]

    K- planes: Explicit radiance fields in space, time, and appear- ance

    Sara Fridovich-Keil, Giacomo Meanti, Frederik Rahbæk Warbæk, Benjamin Recht, and Angjoo Kanazawa. K- planes: Explicit radiance fields in space, time, and appear- ance. InCVPR, 2023. 3

    2023

  19. [27]

    Dynamic view synthesis from dynamic monocu- lar video

    Chen Gao, Ayush Saraf, Johannes Kopf, and Jia-Bin Huang. Dynamic view synthesis from dynamic monocu- lar video. InICCV, 2021. 3

    2021

  20. [28]

    Fluid- nexus: 3d fluid reconstruction and prediction from a single video

    Yue Gao, Hong-Xing Yu, Bo Zhu, and Jiajun Wu. Fluid- nexus: 3d fluid reconstruction and prediction from a single video. InCVPR, 2025. 2, 3 9

    2025

  21. [29]

    Fleet, Saurabh Sax- ena, and Andrea Tagliasacchi

    Lily Goli, Sara Sabour, Mark Matthews, Marcus Brubaker, Dmitry Lagun, Alec Jacobson, David J. Fleet, Saurabh Sax- ena, and Andrea Tagliasacchi. RoMo: Robust motion seg- mentation improves structure from motion.ICCV, 2025. 2, 3

    2025

  22. [30]

    Harley, Zhaoyuan Fang, and Katerina Fragki- adaki

    Adam W. Harley, Zhaoyuan Fang, and Katerina Fragki- adaki. Particle video revisited: Tracking through occlusions using point trajectories. InECCV, 2022. 2

    2022

  23. [31]

    Berthold K. P. Horn and Brian G. Schunck. Determining optical flow.Artificial Intelligence, 1981. 2

    1981

  24. [32]

    2d gaussian splatting for geometrically ac- curate radiance fields

    Binbin Huang, Zehao Yu, Anpei Chen, Andreas Geiger, and Shenghua Gao. 2d gaussian splatting for geometrically ac- curate radiance fields. InSIGGRAPH 2024 Conference Pa- pers, 2024. 3

    2024

  25. [33]

    Sc-gs: Sparse-controlled gaussian splatting for editable dynamic scenes

    Yi-Hua Huang, Yang-Tian Sun, Ziyi Yang, Xiaoyang Lyu, Yan-Pei Cao, and Xiaojuan Qi. Sc-gs: Sparse-controlled gaussian splatting for editable dynamic scenes. InCVPR, pages 4220–4230, 2024. 2, 3

    2024

  26. [34]

    Self-supervised monocular scene flow estimation

    Junhwa Hur and Stefan Roth. Self-supervised monocular scene flow estimation. InCVPR, 2020. 2, 3

    2020

  27. [35]

    Ivo Ihrke and Marcus A. Magnor. Image-based tomo- graphic reconstruction of flames. InProceedings of the ACM SIGGRAPH/Eurographics Symposium on Computer Animation (SCA), pages 365–373, 2004. 2, 3

    2004

  28. [36]

    V olumedeform: Real-time volumetric non-rigid reconstruction

    Matthias Innmann, Michael Zollh ¨ofer, Matthias Nießner, Christian Theobalt, and Marc Stamminger. V olumedeform: Real-time volumetric non-rigid reconstruction. InECCV,

  29. [37]

    Humanrf: High-fidelity neural radiance fields for humans in motion.ACM TOG, 2023

    Mustafa Isik, Martin R ¨unz, Markos Georgopoulos, Taras Khakhulin, Jonathan Starck, Lourdes Agapito, and Matthias Nießner. Humanrf: High-fidelity neural radiance fields for humans in motion.ACM TOG, 2023. 3

    2023

  30. [38]

    Learning to estimate hidden motions with global motion aggregation

    Shihao Jiang, Dylan Campbell, Yao Lu, Hongdong Li, and Richard Hartley. Learning to estimate hidden motions with global motion aggregation. InICCV, 2021. 2

    2021

  31. [39]

    Learning optical flow from a few matches

    Shihao Jiang, Yao Lu, Hongdong Li, and Richard Hartley. Learning optical flow from a few matches. InCVPR, 2021. 2

    2021

  32. [40]

    Co- tracker: It is better to track together

    Nikita Karaev, Ignacio Rocco, Benjamin Graham, Natalia Neverova, Andrea Vedaldi, and Christian Rupprecht. Co- tracker: It is better to track together. InECCV, 2024. 2

    2024

  33. [41]

    3d gaussian splatting for real-time radiance field rendering.ACM Transactions on Graphics (SIGGRAPH), 42(4):1–14, 2023

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk ¨uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering.ACM Transactions on Graphics (SIGGRAPH), 42(4):1–14, 2023. 2, 3, 4, 1

    2023

  34. [42]

    Michels, S¨oren Pirk, and Wojtek Palubicki

    Andrzej Kokosza, Helge Wrede, Daniel Gonzalez Esparza, Milosz Makowski, Daoming Liu, Dominik L. Michels, S¨oren Pirk, and Wojtek Palubicki. Scintilla: Simulating combustible vegetation for wildfires.ACM Trans. Graph., 43(4), 2024. 2

    2024

  35. [43]

    Robust consistent video depth estimation

    Johannes Kopf, Xuejian Rong, and Jia-Bin Huang. Robust consistent video depth estimation. InCVPR, 2021. 3

    2021

  36. [44]

    Tapvid-3d: A benchmark for tracking any point in 3d

    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. InNeurIPS, 2024. 2, 3

    2024

  37. [45]

    DynMF: Neural motion factorization for real-time dynamic view synthesis with 3d gaussian splatting

    Agelos Kratimenos, Jiahui Lei, and Kostas Daniilidis. DynMF: Neural motion factorization for real-time dynamic view synthesis with 3d gaussian splatting. InECCV, 2024. 3

    2024

  38. [46]

    Monoc- ular dense 3d reconstruction of a complex dynamic scene from two perspective frames

    Suryansh Kumar, Yuchao Dai, and Hongdong Li. Monoc- ular dense 3d reconstruction of a complex dynamic scene from two perspective frames. InICCV, 2017. 3

    2017

  39. [47]

    Harley, Leonidas Guibas, and Kostas Daniilidis

    Jiahui Lei, Yijia Weng, Adam W. Harley, Leonidas Guibas, and Kostas Daniilidis. Mosca: Dynamic gaussian fusion from casual videos via 4d motion scaffolds. InCVPR, 2025. 3

    2025

  40. [48]

    Grounding image matching in 3d with mast3r, 2024

    Vincent Leroy, Yohann Cabon, and Jerome Revaud. Grounding image matching in 3d with mast3r, 2024. 2, 3

    2024

  41. [49]

    Tava: Template-free animatable volumetric actors

    Ruilong Li, Julian Tanke, Minh V o, Michael Zollh ¨ofer, J¨urgen Gall, Angjoo Kanazawa, and Christoph Lassner. Tava: Template-free animatable volumetric actors. In ECCV, 2022. 3

    2022

  42. [50]

    Neural 3d video synthesis from multi-view video

    Tianye Li, Mira Slavcheva, Michael Zollhoefer, Simon Green, Christoph Lassner, Changil Kim, Tanner Schmidt, Steven Lovegrove, Michael Goesele, Richard Newcombe, et al. Neural 3d video synthesis from multi-view video. In CVPR, 2022. 3

    2022

  43. [51]

    Zhengqi Li, Tali Dekel, Forrester Cole, Richard Tucker, Noah Snavely, Ce Liu, and William T. Freeman. Learning the depths of moving people by watching frozen people. In CVPR, 2019. 3

    2019

  44. [52]

    Neural scene flow fields for space-time view syn- thesis of dynamic scenes

    Zhengqi Li, Simon Niklaus, Noah Snavely, and Oliver Wang. Neural scene flow fields for space-time view syn- thesis of dynamic scenes. InCVPR, 2021. 3

    2021

  45. [53]

    Dynibar: Neural dynamic image-based rendering

    Zhengqi Li, Qianqian Wang, Forrester Cole, Richard Tucker, and Noah Snavely. Dynibar: Neural dynamic image-based rendering. InCVPR, 2023. 3

    2023

  46. [54]

    Spacetime gaussian feature splatting for real-time dynamic view syn- thesis

    Zhan Li, Zhang Chen, Zhong Li, and Yi Xu. Spacetime gaussian feature splatting for real-time dynamic view syn- thesis. InCVPR, pages 8508–8520, 2024. 3

    2024

  47. [55]

    SIFT flow: Dense correspondence across scenes and its applications

    Ce Liu, Jenny Yuen, and Antonio Torralba. SIFT flow: Dense correspondence across scenes and its applications. IEEE TPAMI, 2011. 2

    2011

  48. [56]

    Modgs: Dy- namic gaussian splatting from casually-captured monocular videos

    Qingming Liu, Yuan Liu, Jiepeng Wang, Xianqiang Lv, Peng Wang, Wenping Wang, and Junhui Hou. Modgs: Dy- namic gaussian splatting from casually-captured monocular videos. InICLR, 2025. 3

    2025

  49. [57]

    David G. Lowe. Distinctive image features from scale- invariant keypoints.IJCV, 2004. 2

    2004

  50. [58]

    Align3r: Aligned monocular depth estimation for dynamic videos

    Jiahao Lu, Tianyu Huang, Peng Li, Zhiyang Dou, Cheng Lin, Zhiming Cui, Zhen Dong, Sai-Kit Yeung, Wenping Wang, and Yuan Liu. Align3r: Aligned monocular depth estimation for dynamic videos. InCVPR, 2025. 3

    2025

  51. [59]

    Lucas and Takeo Kanade

    Bruce D. Lucas and Takeo Kanade. An iterative image registration technique with an application to stereo vision. InInternational Joint Conference on Artificial Intelligence (IJCAI), 1981. 2

    1981

  52. [60]

    Dynamic 3d gaussians: Tracking by per- sistent dynamic view synthesis

    Jonathon Luiten, Georgios Kopanas, Bastian Leibe, and Deva Ramanan. Dynamic 3d gaussians: Tracking by per- sistent dynamic view synthesis. In3DV, pages 800–809,

  53. [61]

    Consistent video depth estimation

    Xuan Luo, Jia-Bin Huang, Richard Szeliski, Kevin Matzen, and Johannes Kopf. Consistent video depth estimation. ACM TOG, 39(4), 2020. 3

    2020

  54. [62]

    Srinivasan, Matthew Tancik, Jonathan T

    Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view syn- thesis. InECCV, pages 405–421. Springer, 2020. 2, 3

    2020

  55. [63]

    MFT: Long- term tracking of every pixel

    Michal Neoral, Jonas Serych, and Jiri Matas. MFT: Long- term tracking of every pixel. InWACV, 2024. 2

    2024

  56. [64]

    Newcombe, Dieter Fox, and Steven M

    Richard A. Newcombe, Dieter Fox, and Steven M. Seitz. Dynamicfusion: Reconstruction and tracking of non-rigid scenes in real-time. InCVPR, 2015. 3

    2015

  57. [65]

    Delta: Dense efficient long-range 3d tracking for any video

    Tuan Duc Ngo, Peiye Zhuang, Chuang Gan, Evange- los Kalogerakis, Sergey Tulyakov, Hsin-Ying Lee, and Chaoyang Wang. Delta: Dense efficient long-range 3d tracking for any video. InICLR, 2025. 2, 3

    2025

  58. [66]

    C3DPO: Canonical 3d pose networks for non-rigid structure from motion

    David Novotny, Nikhila Ravi, Benjamin Graham, Natalia Neverova, and Andrea Vedaldi. C3DPO: Canonical 3d pose networks for non-rigid structure from motion. InICCV,

  59. [67]

    Barron, Sofien Bouaziz, Dan B

    Keunhong Park, Utkarsh Sinha, Jonathan T. Barron, Sofien Bouaziz, Dan B. Goldman, Steven M. Seitz, and Ricardo Martin-Brualla. Nerfies: Deformable neural radiance fields. InICCV, 2021. 3

    2021

  60. [68]

    Barron, Sofien Bouaziz, Dan B

    Keunhong Park, Utkarsh Sinha, Peter Hedman, Jonathan T. Barron, Sofien Bouaziz, Dan B. Goldman, Ricardo Martin- Brualla, and Steven M. Seitz. Hypernerf: A higher- dimensional representation for topologically varying neural radiance fields.ACM TOG, 2021

    2021

  61. [69]

    D-nerf: Neural radiance fields for dynamic scenes

    Albert Pumarola, Enric Corona, Gerard Pons-Moll, and Francesc Moreno-Noguer. D-nerf: Neural radiance fields for dynamic scenes. InCVPR, pages 10318–10327, 2021. 2, 3

    2021

  62. [70]

    Dense monocular depth estimation in complex dy- namic scenes

    Rene Ranftl, Vibhav Vineet, Qifeng Chen, and Vladlen Koltun. Dense monocular depth estimation in complex dy- namic scenes. InCVPR, 2016. 3

    2016

  63. [71]

    Sudderth, and Jan Kautz

    Zhile Ren, Orazio Gallo, Deqing Sun, Ming-Hsuan Yang, Erik B. Sudderth, and Jan Kautz. A fusion approach for multi-frame optical flow estimation. InWACV, 2019. 2

    2019

  64. [72]

    Ethan Rublee, Vincent Rabaud, Kurt Konolige, and Gary R. Bradski. ORB: An efficient alternative to SIFT or SURF. InICCV, 2011. 2

    2011

  65. [73]

    Video pop-up: Monocular 3d reconstruction of dynamic scenes

    Chris Russell, Rui Yu, and Lourdes Agapito. Video pop-up: Monocular 3d reconstruction of dynamic scenes. InECCV,

  66. [74]

    R3d3: Dense 3d reconstruction of dynamic scenes from multiple cameras

    Aron Schmied, Tobias Fischer, Martin Danelljan, Marc Pollefeys, and Fisher Yu. R3d3: Dense 3d reconstruction of dynamic scenes from multiple cameras. InICCV, pages 3216–3226, 2023. 1

    2023

  67. [75]

    Structure-from-motion revisited

    Johannes Lutz Sch ¨onberger and Jan-Michael Frahm. Structure-from-motion revisited. InConference on Com- puter Vision and Pattern Recognition (CVPR), 2016. 2

    2016

  68. [76]

    A vote-and-verify strategy for fast spatial verification in image retrieval

    Johannes Lutz Sch ¨onberger, True Price, Torsten Sattler, Jan-Michael Frahm, and Marc Pollefeys. A vote-and-verify strategy for fast spatial verification in image retrieval. In Asian Conference on Computer Vision (ACCV), 2016. 4

    2016

  69. [77]

    Pixelwise view selection for unstructured multi-view stereo

    Johannes Lutz Sch ¨onberger, Enliang Zheng, Marc Polle- feys, and Jan-Michael Frahm. Pixelwise view selection for unstructured multi-view stereo. InEuropean Conference on Computer Vision (ECCV), 2016. 2

    2016

  70. [78]

    Videoflow: Exploiting temporal cues for multi-frame optical flow estimation

    Xiaoyu Shi, Zhaoyang Huang, Weikang Bian, Dasong Li, Manyuan Zhang, Ka Chun Cheung, Simon See, Hongwei Qin, Jifeng Dai, and Hongsheng Li. Videoflow: Exploiting temporal cues for multi-frame optical flow estimation. In ICCV, 2023. 2

    2023

  71. [79]

    Nerf- player: A streamable dynamic scene representation with decomposed neural radiance fields.IEEE TVCG, 29(5),

    Liangchen Song, Anpei Chen, Zhong Li, Zhang Chen, Lele Chen, Junsong Yuan, Yi Xu, and Andreas Geiger. Nerf- player: A streamable dynamic scene representation with decomposed neural radiance fields.IEEE TVCG, 29(5),

  72. [80]

    Track everything everywhere fast and robustly

    Yunzhou Song, Jiahui Lei, Ziyun Wang, Lingjie Liu, and Kostas Daniilidis. Track everything everywhere fast and robustly. InECCV, 2024. 2

    2024

  73. [81]

    Marigold-dc: Zero-shot monocular depth com- pletion with guided diffusion

    Massimiliano Viola, Kevin Qu, Nando Metzger, Bingxin Ke, Alexander Becker, Konrad Schindler, and Anton Obukhov. Marigold-dc: Zero-shot monocular depth com- pletion with guided diffusion. InICCV, 2025. 2

    2025

  74. [82]

    Neural trajectory fields for dynamic novel view syn- thesis

    Chaoyang Wang, Ben Eckart, Simon Lucey, and Orazio Gallo. Neural trajectory fields for dynamic novel view syn- thesis. InarXiv preprint arXiv:2105.05994, 2021. 3

    2021 arXiv

  75. [83]

    Ode-gs: Latent odes for dynamic scene extrapolation with 3d gaussian splatting

    Daniel Wang, Patrick Rim, Tian Tian, Dong Lao, Alex Wong, and Ganesh Sundaramoorthi. Ode-gs: Latent odes for dynamic scene extrapolation with 3d gaussian splatting. arXiv preprint arXiv:2506.05480, 2025. 3

    2025 arXiv

  76. [84]

    Fourier plenoctrees for dynamic radiance field rendering in real-time

    Liao Wang, Jiakai Zhang, Xinhang Liu, Fuqiang Zhao, Yanshun Zhang, Yingliang Zhang, Minye Wu, Jingyi Yu, and Lan Xu. Fourier plenoctrees for dynamic radiance field rendering in real-time. InCVPR, 2022. 3

    2022

  77. [85]

    Tracking everything everywhere all at once

    Qianqian Wang, Yen-Yu Chang, Ruojin Cai, Zhengqi Li, Bharath Hariharan, Aleksander Holynski, and Noah Snavely. Tracking everything everywhere all at once. In ICCV, 2023. 2

    2023

  78. [86]

    Efros, and Angjoo Kanazawa

    Qianqian Wang, Yifei Zhang, Aleksander Holynski, Alexei A. Efros, and Angjoo Kanazawa. Continuous 3d perception model with persistent state. InCVPR, 2025. 3

    2025

  79. [87]

    Dust3r: Geometric 3d vision made easy

    Shuzhe Wang, Vincent Leroy, Yohann Cabon, Boris Chidlovskii, and Jerome Revaud. Dust3r: Geometric 3d vision made easy. InCVPR, 2024. 2, 3

    2024

  80. [88]

    Freetimegs: Free gaussian primitives at any- time anywhere for dynamic scene reconstruction

    Yifan Wang, Peishan Yang, Zhen Xu, Jiaming Sun, Zhan- hua Zhang, Chen Yong, Hujun Bao, Sida Peng, and Xi- aowei Zhou. Freetimegs: Free gaussian primitives at any- time anywhere for dynamic scene reconstruction. InCVPR,

  81. [89]

    Srinivasan, Jonathan T

    Chung-Yi Weng, Brian Curless, Pratul P. Srinivasan, Jonathan T. Barron, and Ira Kemelmacher-Shlizerman. Hu- mannerf: Free-viewpoint rendering of moving people from monocular video. InCVPR, 2022. 3

    2022

  82. [90]

    Wagner, Sarker Miraz Mahfuz, Wo- jtek Pałubicki, Dominik L

    Helge Wrede, Anton R. Wagner, Sarker Miraz Mahfuz, Wo- jtek Pałubicki, Dominik L. Michels, and S ¨oren Pirk. Fire- X: Extinguishing fire with stoichiometric heat release.ACM TOG, 44(6):1–17, 2025. 2

    2025

  83. [91]

    4d gaussian splatting for real-time dynamic 11 scene rendering

    Guanjun Wu, Taoran Yi, Jiemin Fang, Lingxi Xie, Xi- aopeng Zhang, Wei Wei, Wenyu Liu, Qi Tian, and Xing- gang Wang. 4d gaussian splatting for real-time dynamic 11 scene rendering. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR), pag...

    2024

  84. [92]

    4d gaussian splatting for real-time dynamic scene rendering

    Guanjun Wu, Taoran Yi, Jiemin Fang, Lingxi Xie, Xi- aopeng Zhang, Wei Wei, Wenyu Liu, Qi Tian, and Xing- gang Wang. 4d gaussian splatting for real-time dynamic scene rendering. InCVPR, pages 20310–20320, 2024. 2, 3, 6, 7, 8

    2024

  85. [93]

    3dgut: Enabling dis- torted cameras and secondary rays in gaussian splatting

    Qi Wu, Janick Martinez Esturo, Ashkan Mirzaei, Nicolas Moenne-Loccoz, and Zan Gojcic. 3dgut: Enabling dis- torted cameras and secondary rays in gaussian splatting. CVPR, 2025. 3

    2025

  86. [94]

    Space-time neural irradiance fields for free-viewpoint video

    Wenqi Xian, Jia-Bin Huang, Johannes Kopf, and Changil Kim. Space-time neural irradiance fields for free-viewpoint video. InCVPR, 2021. 3

    2021

  87. [95]

    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. InCVPR, 2024. 2, 3

    2024

  88. [96]

    Spatialtrackerv2: 3d point tracking made easy

    Yuxi Xiao, Jianyuan Wang, Nan Xue, Nikita Karaev, Yuri Makarov, Bingyi Kang, Xing Zhu, Hujun Bao, Yujun Shen, and Xiaowei Zhou. Spatialtrackerv2: 3d point tracking made easy. InICCV, 2025

    2025

  89. [97]

    4dgt: Learning a 4d gaus- sian transformer using real-world monocular videos

    Zhen Xu, Zhengqin Li, Zhao Dong, Xiaowei Zhou, Richard Newcombe, and Zhaoyang Lv. 4dgt: Learning a 4d gaus- sian transformer using real-world monocular videos. In NeurIPS, 2025. 2, 3

    2025

  90. [98]

    Freeman, and Ce Liu

    Gengshan Yang, Deqing Sun, Varun Jampani, Daniel Vlasic, Forrester Cole, Huiwen Chang, Deva Ramanan, William T. Freeman, and Ce Liu. Lasr: Learning artic- ulated shape reconstruction from a monocular video. In CVPR, 2021. 3

    2021

  91. [99]

    Viser: Video-specific surface embeddings for articulated 3d shape reconstruction

    Gengshan Yang, Deqing Sun, Varun Jampani, Daniel Vla- sic, Forrester Cole, Ce Liu, and Deva Ramanan. Viser: Video-specific surface embeddings for articulated 3d shape reconstruction. InNeurIPS, 2021

    2021

  92. [100]

    Banmo: Build- ing animatable 3d neural models from many casual videos

    Gengshan Yang, Minh V o, Natalia Neverova, Deva Ra- manan, Andrea Vedaldi, and Hanbyul Joo. Banmo: Build- ing animatable 3d neural models from many casual videos. InCVPR, 2022. 3

    2022

  93. [101]

    Depth anything: Un- leashing the power of large-scale unlabeled data

    Lihe Yang, Bingyi Kang, Zilong Huang, Xiaogang Xu, Ji- ashi Feng, and Hengshuang Zhao. Depth anything: Un- leashing the power of large-scale unlabeled data. InCVPR,

  94. [102]

    Depth any- thing v2

    Lihe Yang, Bingyi Kang, Zilong Huang, Zhen Zhao, Xiao- gang Xu, Jiashi Feng, and Hengshuang Zhao. Depth any- thing v2. InNeurIPS, Red Hook, NY , USA, 2024. Curran Associates Inc. 2, 4, 7

    2024

  95. [103]

    Deformable 3d gaussians for high-fidelity monocular dynamic scene reconstruction

    Ziyi Yang, Xinyu Gao, Wen Zhou, Shaohui Jiao, Yuqing Zhang, and Xiaogang Jin. Deformable 3d gaussians for high-fidelity monocular dynamic scene reconstruction. In CVPR, 2024. 3

    2024

  96. [104]

    Real- time photorealistic dynamic scene representation and ren- dering with 4d gaussian splatting

    Zeyu Yang, Hongye Yang, Zijie Pan, and Li Zhang. Real- time photorealistic dynamic scene representation and ren- dering with 4d gaussian splatting. InICLR, 2024. 3, 6, 7, 8, 1, 4

    2024

  97. [105]

    Novel view synthesis of dynamic scenes with globally coherent depths from a monocular camera

    Jae Shin Yoon, Kihwan Kim, Orazio Gallo, Hyun Soo Park, and Jan Kautz. Novel view synthesis of dynamic scenes with globally coherent depths from a monocular camera. In CVPR, 2020. 3

    2020

  98. [106]

    Freeman, and Tali Dekel

    Zhoutong Zhang, Forrester Cole, Richard Tucker, William T. Freeman, and Tali Dekel. Consistent depth of moving objects in video.ACM TOG, 40(4), 2021. 3

    2021

  99. [107]

    Zhoutong Zhang, Forrester Cole, Zhengqi Li, Michael Ru- binstein, Noah Snavely, and William T. Freeman. Structure and motion from casual videos. InECCV, 2022. 3

    2022

  100. [108]

    Harley, Bokui Shen, Gordon Wet- zstein, and Leonidas J

    Yang Zheng, Adam W. Harley, Bokui Shen, Gordon Wet- zstein, and Leonidas J. Guibas. Pointodyssey: A large-scale synthetic dataset for long-term point tracking. InICCV,

  101. [109]

    Loop, Christian Theobalt, and Marc Stamminger

    Michael Zollh ¨ofer, Matthias Nießner, Shahram Izadi, Christoph Rhemann, Christopher Zach, Matthew Fisher, Chenglei Wu, Andrew William Fitzgibbon, Charles T. Loop, Christian Theobalt, and Marc Stamminger. Real-time non-rigid reconstruction using an rgb-d camera.ACM TOG, 33, 20...

    2014

  102. [110]

    In the video, we show the rendered results for different kinds of trajecto- ries, leaving the camera poses used for training

    Novel View Trajectories To provide a qualitative demonstration of the capability of our model to generalize to novel views, we provide the videoresults and pipeline.mp4. In the video, we show the rendered results for different kinds of trajecto- ries, leaving the camera poses ...

  103. [111]

    For one scene, this video shows the relevant preprocessing and op- timization steps

    Pipeline Visualization In the second half of the same video file we present a visu- alization of our proposed reconstruction pipeline. For one scene, this video shows the relevant preprocessing and op- timization steps. For each stage, we show the correspond- ing view for each...

  104. [112]

    To account for the undercon- strained scene geometry, we increased the depth loss by a factor of100

    Implementation Details For the static scene optimization, we applied the original 3DGS implementation [41]. To account for the undercon- strained scene geometry, we increased the depth loss by a factor of100. To prevent the introduction of floating ar- tifacts in the backgroun...

  105. [113]

    Due to size limitations of the supplementary ma- terial, we provide only a single short example scene in the datasubdirectory

    Code We include the Python source code for preprocessing, as well as for the static and dynamic optimization, in thecode directory. Due to size limitations of the supplementary ma- terial, we provide only a single short example scene in the datasubdirectory. The source code co...

  106. [114]

    With this setup, training a single scene takes approxi- mately30 minat a GPU memory usage of8 GBand a sys- tem memory usage of2 GB

    Compute Hardware For training, we used a system with a 16-core AMD Ryzen Threadripper PRO 3955, an NVIDIA RTX A6000 with 48 GBof GPU memory, as well as128 GBof system mem- ory. With this setup, training a single scene takes approxi- mately30 minat a GPU memory usage of8 GBand ...

  107. [115]

    We provide a more detailed overview of all captured scene in Tab

    Dataset Description For the validation of our proposed method, we captured a diverse dataset of different fuels and backgrounds. We provide a more detailed overview of all captured scene in Tab. 3. Each video has a duration of15 s, a resolution of 1280×720, and a frame rate of400 Hz

  108. [116]

    5 only reports the metrics averaged over all scenes, we provide the more detailed per-scene results of our pro- posed method in Tab

    Per-Scene Results As Sec. 5 only reports the metrics averaged over all scenes, we provide the more detailed per-scene results of our pro- posed method in Tab. 4. The per-scene results for the two baselines we compared our method against are shown in Tabs. 5 and 6, whereas the ...

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.