Pith. sign in

REVIEW 4 major objections 6 minor 64 references

DATAP-SfM: Dynamic-Aware Tracking Any Point for Robust Structure from Motion in the Wild

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a dynamic-aware point tracking network, combined with consistent video depth and global bundle adjustment over static visible tracks, achieves state-of-the-art camera pose estimation in dynamic scenes, with ATE…

desk verdict A genuinely useful dynamic point tracker for SfM, but the headline numbers mix two different depth configurations and need to be fixed before the SOTA claims can be taken at face value. read the letter →

arxiv 2411.13291 v1 pith:VZSWUM3J submitted 2024-11-20 cs.CV

Weicai Ye , Xinyu Chen , Ruohao Zhan , Di Huang , Xiaoshui Huang , Haoyi Zhu , Hujun Bao , Wanli Ouyang
show 2 more authors
Tong He Guofeng Zhang
This is my paper · ORCID
classification cs.CV
keywords dynamic-awarepointtrackingstructurefrommotionsegmentationconsistentvideodepthcameraposeestimationglobalbundleadjustment
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper is trying to establish that dynamic-aware dense point tracking, built on long-term video correspondence and consistent video depth, can replace the chain of adjacent-frame optical flow that previous trajectory-based SfM methods rely on. If correct, this removes the cumulative error of chaining pairwise matches and the scale ambiguity of single-view depths, and lets all camera poses be optimized at once over point tracks that are classified as static and visible. The practical payoff is reliable camera trajectory estimation and dense reconstruction from casual monocular videos containing people, vehicles, and other moving objects, where classic SfM and SLAM typically fail.

What carries the argument

The load-bearing object is the depth-aware trajectory motion segmentation module inside DATAP. It takes per-frame monocular or video depth, normalizes each depth map to (0,1), back-projects the 2D point tracks into 3D camera coordinates, and concatenates 2D trajectory positions, 3D positions, 2D motion, and scene-flow motion into a $10L$-dimensional feature vector per trajectory. These features are encoded by a transformer and decoded by an OANet-style network with diff-pool clustering to predict dynamic labels. This module is what disambiguates static from moving points; the ablations show removing it degrades mIoU from 54.9 to 49.8 and full-set ATE from 0.104 to 0.138.

What would settle it

Run the pipeline on dynamic sequences with ground-truth depth but per-frame scale perturbations that keep normalized depth in (0,1); if motion segmentation mIoU and ATE degrade significantly, the assumption that normalized depth suffices is wrong. Alternatively, substituting a constant depth map while keeping everything else fixed should reproduce the 'w/o depth' ablation numbers; if it does not, the depth module is doing something else.

Watch

Extended reading notes

Core claim

The central claim is that a single sliding-window transformer, which jointly estimates point trajectories, visibility, and per-point dynamic labels, plus global bundle adjustment over static visible tracks, yields state-of-the-art camera pose estimation on dynamic sequences. The paper reports a 19.37% improvement in ATE over ParticleSfM on the full MPI Sintel set and a 26.62% improvement on the TUM RGBD dynamic subset, and shows that adding depth priors lifts motion segmentation mIoU from 49.8 to 54.9. The paper also states that ParticleSfM fails outright on some TUM sequences while the proposed method returns a pose.

Load-bearing premise

The method assumes that normalizing per-frame depth to (0,1) and back-projecting 2D tracks into 3D preserves enough scene-flow signal to tell static from moving points, even though per-frame depth estimates carry unknown scale.

Editorial extensions

If this is right

  • Global structure from motion can run directly on long-term point tracks, skipping incremental camera registration and its drift.
  • Consistent video depth improves motion segmentation enough to matter downstream: mIoU goes from 49.8 to 54.9 in the paper's ablations.
  • The pipeline produces complete point clouds of dynamic scenes by fusing depth-based dynamic points into the static SfM point cloud.
  • The method can serve as an initialization for dynamic-scene NeRF reconstruction and editing, as the paper notes in its broader-impact section.

Reading between the lines

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

  • If the normalized-depth assumption holds, the same architecture should transfer to any video where relative scene flow is informative, even without metric scale; a testable variant would replace depth with a fixed scale and measure the mIoU drop.
  • The dependence on a learned video-depth model (NVDS) may be the main bottleneck for wild videos; a fully self-supervised depth cue trained jointly with tracking could make the pipeline more general.
  • The global-bundle-adjustment design suggests that robustness gains should grow with sequence length, because no incremental registration error accumulates; this could be checked by scaling experiments on longer videos.
  • If the method is made efficient enough for real-time use, it could form the front end of a dynamic SLAM system; the paper flags this direction as future work.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes DATAP-SfM, a structure-from-motion pipeline for monocular videos containing dynamic objects. The core contribution is a dynamic-aware point-tracking network (DATAP) that jointly estimates long-term 2D trajectories, visibility, and per-point dynamic labels in a sliding-window transformer, aided by a depth-aware motion-segmentation decoder. The resulting static, visible tracks are fed into global bundle adjustment to estimate camera poses and a dense point cloud. Experiments on MPI Sintel, TUM RGBD dynamic sequences, and DAVIS compare against ParticleSfM, DROID-SLAM, COLMAP, and others, reporting state-of-the-art ATE improvements. The paper also provides ablations on the depth source and motion-segmentation components.

Significance. If the reported results are reproducible, the paper makes a useful contribution: it integrates modern long-term point tracking with learned dynamic segmentation to enable global, rather than incremental, SfM in dynamic scenes, and it evaluates on standard benchmarks with several baselines and ablations. The strengths include the use of held-out benchmarks (MPI Sintel, TUM RGBD), a transparent training protocol (FlyingThings3D, 4 GPUs, 2 days), and a clear decomposition of the pipeline into tracking, motion segmentation, and bundle adjustment. However, the central quantitative claim of state-of-the-art pose estimation is currently undermined by an inconsistency between Table 2 and the ablation results in Table 4, as detailed below. The paper also does not discuss code or model release, which limits reproducibility. Overall, the approach is plausible and the experimental design is appropriate, but the headline numbers need to be reconciled with a single, clearly specified configuration before the claims can be accepted.

major comments (4)
  1. [Table 2 vs. Table 4] The 'Ours' row in Table 2 does not correspond to any single ablation variant in Table 4. On the MPI Sintel full set, Table 2 lists ATE 0.104, RPE Trans 0.037, RPE Rot 0.306, which exactly match the 'w/ depth' variant in Table 4; on the COLMAP subset, Table 2 lists ATE 0.015, RPE Trans 0.007, RPE Rot 0.128, which exactly match the 'w/ consistent depth' variant. The paper never states that different depth sources are used for the two subsets, so the 'Ours' row is not reproducible as a single pipeline. This directly affects the headline claims of a 19.37% improvement on the full set and a 28.57% improvement on the COLMAP subset. Please either report one designated configuration for Table 2 or explicitly present the per-subset depth-source choices and justify why such mixing is a fair comparison.
  2. [Sections 1 and 3.1; Table 4] The abstract and method repeatedly emphasize consistent video depth (NVDS) as the solution to scale ambiguity and as a performance booster, but Table 4 shows that the monocular per-frame depth variant ('w/ depth') achieves a better full-set ATE (0.104) than the consistent-depth variant (0.117). The only advantage of consistent depth appears on the COLMAP subset (ATE 0.015 vs. 0.023). The paper does not discuss this discrepancy or explain why the advocated depth source is worse on the full set. This is a load-bearing point because the central pose-estimation claim is tied to the full-set number. Please provide an analysis of this trade-off or qualify the claims about consistent depth accordingly.
  3. [Section 3.1, depth normalization] The depth-aware motion segmentation normalizes per-frame monocular or video depth to (0,1) and back-projects 2D tracks to 3D to construct scene-flow features. Since single-view and video depth estimates are defined only up to an unknown per-frame scale, the absolute scale of the 3D trajectory features is not meaningful, and the normalization may distort the relative motion cue. The paper asserts that consistent video depth alleviates this ambiguity, but the ablations in Table 4 do not support a consistent advantage, and no scale-invariant alternative is tested. Please provide a justification for why this normalization preserves the static-versus-dynamic signal, or add an ablation with scale-invariant features (e.g., normalized scene flow with per-trajectory scaling).
  4. [Table 3 and Section 4.5] The TUM RGBD evaluation is not clearly specified. The table header uses contradictory row labels ('Full set' followed by 'Full subset'), the same 'Ours' method appears in both the ParticleSfM subset and full-set rows with different numbers, and the text does not list which 9 sequences were selected or define the 'ParticleSfM subset.' Without this information, the claimed 26.62% improvement on the subset cannot be independently verified. Please provide the sequence list, the exact evaluation protocol, and a clarified table.
minor comments (6)
  1. [Title and abstract] The title uses 'DATAP-SfM' while the manuscript header and abstract use 'DA TAP-SfM'; please use one consistent spelling throughout.
  2. [Table 4] The header '(Left: Full set, Right: Successful subset of COLMAP)' is misleading because each cell contains two numbers separated by a slash; split the table into two sub-tables or add explicit column headers such as 'Full set / COLMAP subset.'
  3. [Section 3.1, Eq. (1)] The trajectory loss is written as a sum over J of ||P(j) - P_hat(j)||, but the relationship between the index j, the number of trajectories N, and the sliding-window length L is not defined; please clarify the notation.
  4. [Section 4.5] The paper does not list the 9 selected TUM RGBD sequences; including the sequence names would improve reproducibility, as would stating the exact subset definition used for the 'successful' rows.
  5. [Throughout] The paper does not mention whether code or trained models will be released; given the reliance on external components (CoTracker, MiDaS, NVDS) and the difficulty of reproducing the full pipeline, a code-release statement would be valuable.
  6. [References] The bibliography contains duplicate entries for Schonberger & Frahm 2016 and Butler et al. 2012a/b with inconsistent formatting; please unify citations.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation found; the SOTA claim is empirical on external benchmarks, with a table-reporting inconsistency that is not circularity.

full rationale

The paper's central claim is an empirically evaluated pipeline: DATAP is trained on FlyingThings3D with ground-truth trajectory/visibility/dynamic labels, and pose accuracy is measured on held-out MPI Sintel and TUM RGBD sequences against COLMAP, ParticleSfM, DROID-SLAM, and others. No equation in the paper reduces a predicted quantity to a fitted parameter by construction, and no load-bearing argument is supported solely by a self-citation. The depth-aware dynamic segmentation module consumes consistent or monocular depth as an external prior and is supervised by ground-truth dynamic labels; its ablations show it helps, but the improvement is not a definitional or fitting artifact. The self-citations that do appear (Ye et al. 2022, 2023; Zhang et al. 2009; Tan et al. 2013) are related-work mentions and are not load-bearing for the proposed derivation. One reporting concern is worth noting but belongs to correctness/reproducibility rather than circularity: Table 2's 'Ours' row reports full-set ATE 0.104 and subset ATE 0.015, while Table 4 shows the full-set 0.104 comes from the per-frame-depth variant and the subset 0.015 comes from the consistent-depth variant, so the headline row combines the best of two configurations. This weakens the '19.37% improvement' claim as stated, but it does not make any prediction equivalent to its inputs. The paper even includes a limitation section acknowledging the system is not real-time, which is unrelated to circularity. Overall, the derivation is self-contained against external benchmarks, and no significant circularity is present.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central contribution is empirical. The pipeline rests on a set of hand-set hyperparameters (window length, iteration count, loss weights, depth normalization, training schedule) and on domain assumptions about the depth prior and training transfer. No new physical entities are introduced; the only new artifact is a learned neural network module (dynamic prediction heads on a point tracker).

free parameters (5)
  • per-frame depth normalization range = (0,1)
    Each frame's relative depth is normalized to (0,1) before back-projection to 3D; this hand-set scale determines the scene-flow magnitude and is not derived from physical scale. It is central to the dynamic segmentation module.
  • sliding window length L
    The length of the tracking window is a hand-chosen hyperparameter that affects trajectory consistency and computation; its value is not reported in the paper.
  • transformer iteration count K
    The number of iterative transformer updates is a hand-chosen hyperparameter; the value is not reported.
  • loss weights (lambda_traj, lambda_vis, lambda_dyn)
    Weights that combine trajectory, visibility, and dynamic cross-entropy losses are chosen by hand; values are not given, despite being part of the training procedure that produces the final model.
  • training schedule (epochs and learning rate) = 50/30/30 epochs; lr=5e-4
    The two-stage training schedule and learning rate are hand-set; they affect the final model's performance and are not derived from any first-principles argument.
assumptions (4)
  • domain assumption Normalizing per-frame depth to (0,1) preserves enough scene-flow signal to separate dynamic and static points.
    Invoked in Section 3.1 'Depth-aware trajectory motion segmentation'. The ablation shows the method degrades without depth, but the paper never resolves the scale ambiguity of monocular or video depth prior.
  • domain assumption The dynamic label for each track's starting point is initialized to 1 and corrected by the network.
    Section 3.1 states 'the state of the starting point of each track is visible ... and dynamic ... = 1'. This initialization assumes the network can correct an initial all-dynamic bias, which is not analyzed.
  • domain assumption Training on FlyingThings3D generalizes to MPI Sintel, TUM RGBD, and DAVIS.
    Section 4.1 trains only on FlyingThings3D; the paper assumes the learned dynamic-aware tracking transfers to other synthetic and real dynamic scenes.
  • standard math Given accurate static/visible point tracks, global bundle adjustment yields correct camera poses.
    This is the standard SfM assumption underlying the pipeline in Section 3.2; it is accepted practice and not a novel postulate.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DATAP-SfM: Dynamic-Aware Tracking Any Point for Robust Structure from Motion in the Wild." pith.science (2026). https://pith.science/paper/VZSWUM3J

@misc{pith2026241113291,
  author       = {Pith},
  title        = {Pith review of: DATAP-SfM: Dynamic-Aware Tracking Any Point for Robust Structure from Motion in the Wild},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VZSWUM3J}},
  note         = {Machine review of arXiv:2411.13291}
}
read the original abstract

This paper proposes a concise, elegant, and robust pipeline to estimate smooth camera trajectories and obtain dense point clouds for casual videos in the wild. Traditional frameworks, such as ParticleSfM~\cite{zhao2022particlesfm}, address this problem by sequentially computing the optical flow between adjacent frames to obtain point trajectories. They then remove dynamic trajectories through motion segmentation and perform global bundle adjustment. However, the process of estimating optical flow between two adjacent frames and chaining the matches can introduce cumulative errors. Additionally, motion segmentation combined with single-view depth estimation often faces challenges related to scale ambiguity. To tackle these challenges, we propose a dynamic-aware tracking any point (DATAP) method that leverages consistent video depth and point tracking. Specifically, our DATAP addresses these issues by estimating dense point tracking across the video sequence and predicting the visibility and dynamics of each point. By incorporating the consistent video depth prior, the performance of motion segmentation is enhanced. With the integration of DATAP, it becomes possible to estimate and optimize all camera poses simultaneously by performing global bundle adjustments for point tracking classified as static and visible, rather than relying on incremental camera registration. Extensive experiments on dynamic sequences, e.g., Sintel and TUM RGBD dynamic sequences, and on the wild video, e.g., DAVIS, demonstrate that the proposed method achieves state-of-the-art performance in terms of camera pose estimation even in complex dynamic challenge scenes.

Figures

Figures reproduced from arXiv: 2411.13291 by the authors.

Figure 1
Figure 1. Given casual videos, our method can obtain smooth camera trajectories and entire point clouds of dynamic scenes. From top to bottom: video samples, results from COLMAP, ParticleSfM, and ours. Abstract This paper proposes a concise, elegant, and ro￾bust pipeline to estimate smooth camera trajec￾tories and obtain dense point clouds for casual videos in the wild. Traditional frameworks, such as ParticleSfM (Zhao et al.… view at source ↗
Figure 2
Figure 2. DATAP-SfM pipeline. Given monocular videos as input with consistent video depth (optional), DATAP can simultaneously estimate long-term point tracking with visible and dynamic characteristics. Incorporating with DATAP, we formalize a concise, elegant, and robust pipeline of structure from motion by performing global bundle adjustment for point tracking classified as static and visible. impressive results. However, s… view at source ↗
Figure 3
Figure 3. Qualitative results of motion segmentation on MPI Sintel dataset. Our method outperforms existing SOTA methods. From top to bottom: image samples, motion segmentation results from Oneformer (Jain et al., 2023), ParticleSfM, and ours. Red: static, green: dynamic. The third column is the sleeping case, which should be static [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Qualitative results of camera pose estimation on MPI Sintel dataset. Our method outperforms existing SOTA methods [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Qualitative results of structure from motion on DAVIS dataset. Our method can obtain smooth camera trajectories and entire point clouds of dynamic scenes. ation metrics for motion segmentation, such as prediction, recall, F1 score, and IoU. It is worth noting that in t…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

64 extracted references · 54 canonical work pages

  1. [1]

    M., Civera, J., and Neira, J

    Bescos, B., F \'a cil, J. M., Civera, J., and Neira, J. Dynaslam: Tracking, mapping, and inpainting in dynamic scenes. IEEE Robotics and Automation Letters, 3 0 (4): 0 4076--4083, 2018

    work page 2018

  2. [2]

    Midas v3.1 -- a model zoo for robust monocular relative depth estimation

    Birkl, R., Wofk, D., and M \"u ller, M. Midas v3.1 -- a model zoo for robust monocular relative depth estimation. arXiv preprint arXiv:2307.14460, 2023

    arXiv 2023

  3. [3]

    and Malik, J

    Brox, T. and Malik, J. Large displacement optical flow: descriptor matching in variational motion estimation. IEEE transactions on pattern analysis and machine intelligence, 33 0 (3): 0 500--513, 2010

    work page 2010

  4. [4]

    Variational motion segmentation with level sets

    Brox, T., Bruhn, A., and Weickert, J. Variational motion segmentation with level sets. In Computer Vision--ECCV 2006: 9th European Conference on Computer Vision, Graz, Austria, May 7-13, 2006. Proceedings, Part I 9, pp.\ 471--483. Springer, 2006

    work page 2006

  5. [5]

    J., Wulff, J., Stanley, G

    Butler, D. J., Wulff, J., Stanley, G. B., and Black, M. J. A naturalistic open source movie for optical flow evaluation ( S intel movie copyright B lender F oundation, www.sintel.org). In ECCV, 2012 a

    work page 2012

  6. [6]

    J., Wulff, J., Stanley, G

    Butler, D. J., Wulff, J., Stanley, G. B., and Black, M. J. A naturalistic open source movie for optical flow evaluation. In A. Fitzgibbon et al. (Eds.) (ed.), European Conf. on Computer Vision (ECCV), Part IV, LNCS 7577, pp.\ 611--625. Springer-Verlag, October 2012 b

    work page 2012

  7. [7]

    Campos, C., Elvira, R., Rodr \' guez, J. J. G., Montiel, J. M., and Tard \'o s, J. D. Orb-slam3: An accurate open-source library for visual, visual-inertial and multi-map slam. arXiv preprint arXiv:2007.11898, 2020

    arXiv 2007

  8. [8]

    and Govindu, V

    Chatterjee, A. and Govindu, V. M. Efficient and robust large-scale rotation averaging. In Proceedings of the IEEE International Conference on Computer Vision, pp.\ 521--528, 2013

    work page 2013

Show all 64 references
  1. [9]

    and Tan, P

    Cui, Z. and Tan, P. Global structure-from-motion by similarity averaging. In Proceedings of the IEEE International Conference on Computer Vision, pp.\ 864--872, 2015

    2015

  2. [10]

    X., Savva, M., Halber, M., Funkhouser, T., and Nie ner, M

    Dai, A., Chang, A. X., Savva, M., Halber, M., Funkhouser, T., and Nie ner, M. Scannet: Richly-annotated 3d reconstructions of indoor scenes. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp.\ 5828--5839, 2017

    2017

  3. [11]

    Superpoint: Self-supervised interest point detection and description

    DeTone, D., Malisiewicz, T., and Rabinovich, A. Superpoint: Self-supervised interest point detection and description. In Proceedings of the IEEE conference on computer vision and pattern recognition workshops, pp.\ 224--236, 2018

    2018

  4. [12]

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

    Doersch, C., Yang, Y., Vecerik, M., Gokay, D., Gupta, A., Aytar, Y., Carreira, J., and Zisserman, A. Tapir: Tracking any point with per-frame initialization and temporal refinement. ICCV, 2023

    2023

  5. [13]

    and Fergus, R

    Eigen, D. and Fergus, R. Predicting depth, surface normals and semantic labels with a common multi-scale convolutional architecture. In Proceedings of the IEEE international conference on computer vision, pp.\ 2650--2658, 2015

    2015

  6. [14]

    Depth map prediction from a single image using a multi-scale deep network

    Eigen, D., Puhrsch, C., and Fergus, R. Depth map prediction from a single image using a multi-scale deep network. In NeurIPS, 2014

    2014

  7. [15]

    Lsd-slam: Large-scale direct monocular slam

    Engel, J., Sch \"o ps, T., and Cremers, D. Lsd-slam: Large-scale direct monocular slam. In European conference on computer vision, pp.\ 834--849. Springer, 2014

    2014

  8. [16]

    Direct sparse odometry

    Engel, J., Koltun, V., and Cremers, D. Direct sparse odometry. IEEE transactions on pattern analysis and machine intelligence, 40 0 (3): 0 611--625, 2017

    2017

  9. [17]

    Svo: Fast semi-direct monocular visual odometry

    Forster, C., Pizzoli, M., and Scaramuzza, D. Svo: Fast semi-direct monocular visual odometry. In ICRA, pp.\ 15--22. IEEE, 2014

    2014

  10. [18]

    Svo: Semidirect visual odometry for monocular and multicamera systems

    Forster, C., Zhang, Z., Gassner, M., Werlberger, M., and Scaramuzza, D. Svo: Semidirect visual odometry for monocular and multicamera systems. IEEE Transactions on Robotics, 33 0 (2): 0 249--265, 2016

    2016

  11. [19]

    evo: Python package for the evaluation of odometry and slam

    Grupp, M. evo: Python package for the evaluation of odometry and slam. https://github.com/MichaelGrupp/evo, 2017

    2017

  12. [20]

    W., Fang, Z., and Fragkiadaki, K

    Harley, A. W., Fang, Z., and Fragkiadaki, K. Particle video revisited: Tracking through occlusions using point trajectories. In ECCV, 2022

    2022

  13. [21]

    Mask r-cnn

    He, K., Gkioxari, G., Dollar, P., and Girshick, R. Mask r-cnn. In Proceedings of the IEEE International Conference on Computer Vision (ICCV), Oct 2017

    2017

  14. [22]

    Flownet 2.0: Evolution of optical flow estimation with deep networks

    Ilg, E., Mayer, N., Saikia, T., Keuper, M., Dosovitskiy, A., and Brox, T. Flownet 2.0: Evolution of optical flow estimation with deep networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp.\ 2462--2470, 2017

    2017

  15. [23]

    T., Hassani, A., Orlov, N., and Shi, H

    Jain, J., Li, J., Chiu, M. T., Hassani, A., Orlov, N., and Shi, H. Oneformer: One transformer to rule universal image segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 2989--2998, 2023

    2023

  16. [24]

    CoTracker : It is better to track together

    Karaev, N., Rocco, I., Graham, B., Neverova, N., Vedaldi, A., and Rupprecht, C. CoTracker : It is better to track together. 2023

    2023

  17. [25]

    Simvodis: Simultaneous visual odometry, object detection, and instance segmentation, 2019

    Kim, U.-H., Kim, S.-H., and Kim, J.-H. Simvodis: Simultaneous visual odometry, object detection, and instance segmentation, 2019

    2019

  18. [26]

    and Murray, D

    Klein, G. and Murray, D. Parallel tracking and mapping for small ar workspaces. In 2007 6th IEEE and ACM International Symposium on Mixed and Augmented Reality, pp.\ 225--234, 2007. doi:10.1109/ISMAR.2007.4538852

    2007

  19. [27]

    Robust consistent video depth estimation

    Kopf, J., Rong, X., and Huang, J.-B. Robust consistent video depth estimation. arXiv preprint arXiv:2012.05901, 2020

    2012 arXiv

  20. [28]

    Robust consistent video depth estimation

    Kopf, J., Rong, X., and Huang, J.-B. Robust consistent video depth estimation. In IEEE\/CVF Conference on Computer Vision and Pattern Recognition, 2021

    2021

  21. [29]

    Unsupervised Joint Learning of Depth, Optical Flow, Ego-motion from Video

    Li, J., Zhao, J., Song, S., and Feng, T. Unsupervised Joint Learning of Depth, Optical Flow, Ego-motion from Video . arXiv preprint arXiv:2105.14520, 2021

    2021 arXiv

  22. [30]

    and Snavely, N

    Li, Z. and Snavely, N. Megadepth: Learning single-view depth prediction from internet photos. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2018

    2018

  23. [31]

    The emergence of objectness: Learning zero-shot segmentation from videos

    Liu, R., Wu, Z., Yu, S., and Lin, S. The emergence of objectness: Learning zero-shot segmentation from videos. Advances in Neural Information Processing Systems, 34: 0 13137--13152, 2021

    2021

  24. [32]

    Zero-shot video object segmentation with co-attention siamese networks

    Lu, X., Wang, W., Shen, J., Crandall, D., and Luo, J. Zero-shot video object segmentation with co-attention siamese networks. IEEE transactions on pattern analysis and machine intelligence, 44 0 (4): 0 2228--2242, 2020

    2020

  25. [33]

    Consistent video depth estimation

    Luo, X., Huang, J.-B., Szeliski, R., Matzen, K., and Kopf, J. Consistent video depth estimation. ACM Transactions on Graphics (TOG), 39 0 (4): 0 71--1, 2020

    2020

  26. [34]

    and Tard \'o s, J

    Mur-Artal, R. and Tard \'o s, J. D. Orb-slam2: An open-source slam system for monocular, stereo, and rgb-d cameras. IEEE Transactions on Robotics, 33 0 (5): 0 1255--1262, 2017

    2017

  27. [35]

    A., Lovegrove, S

    Newcombe, R. A., Lovegrove, S. J., and Davison, A. J. Dtam: Dense tracking and mapping in real-time. In 2011 international conference on computer vision, pp.\ 2320--2327. IEEE, 2011

    2011

  28. [36]

    and Singer, A

    Ozyesil, O. and Singer, A. Robust camera location estimation by convex programming. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp.\ 2674--2683, 2015

    2015

  29. [37]

    A benchmark dataset and evaluation methodology for video object segmentation

    Perazzi, F., Pont-Tuset, J., McWilliams, B., Van Gool, L., Gross, M., and Sorkine-Hornung, A. A benchmark dataset and evaluation methodology for video object segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp.\ 724--732, 2016

    2016

  30. [38]

    Towards robust monocular depth estimation: Mixing datasets for zero-shot cross-dataset transfer

    Ranftl, R., Lasinger, K., Hafner, D., Schindler, K., and Koltun, V. Towards robust monocular depth estimation: Mixing datasets for zero-shot cross-dataset transfer. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44 0 (3), 2022

    2022

  31. [39]

    Superglue: Learning feature matching with graph neural networks

    Sarlin, P.-E., DeTone, D., Malisiewicz, T., and Rabinovich, A. Superglue: Learning feature matching with graph neural networks. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp.\ 4938--4947, 2020

    2020

  32. [40]

    Schonberger, J. L. and Frahm, J.-M. Structure-from-motion revisited. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp.\ 4104--4113, 2016

    2016

  33. [41]

    Sch\" o nberger, J. L. and Frahm, J.-M. Structure-from-motion revisited. In Conference on Computer Vision and Pattern Recognition (CVPR), 2016

    2016

  34. [42]

    and Malik, J

    Shi, J. and Malik, J. Motion segmentation and tracking using normalized cuts. In Sixth international conference on computer vision (IEEE Cat. No. 98CH36271), pp.\ 1154--1160. IEEE, 1998

    1998

  35. [43]

    A benchmark for the evaluation of rgb-d slam systems

    Sturm, J., Engelhard, N., Endres, F., Burgard, W., and Cremers, D. A benchmark for the evaluation of rgb-d slam systems. In 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems, pp.\ 573--580. IEEE, 2012

    2012

  36. [44]

    Theia: A fast and scalable structure-from-motion library

    Sweeney, C., Hollerer, T., and Turk, M. Theia: A fast and scalable structure-from-motion library. In Proceedings of the 23rd ACM international conference on Multimedia, pp.\ 693--696, 2015

    2015

  37. [45]

    Robust monocular slam in dynamic environments

    Tan, W., Liu, H., Dong, Z., Zhang, G., and Bao, H. Robust monocular slam in dynamic environments. In 2013 IEEE International Symposium on Mixed and Augmented Reality (ISMAR), pp.\ 209--218, 2013. doi:10.1109/ISMAR.2013.6671781

    2013

  38. [46]

    and Deng, J

    Teed, Z. and Deng, J. Raft: Recurrent all-pairs field transforms for optical flow. In European Conference on Computer Vision, pp.\ 402--419. Springer, 2020 a

    2020

  39. [47]

    and Deng, J

    Teed, Z. and Deng, J. Raft: Recurrent all-pairs field transforms for optical flow. In European conference on computer vision, pp.\ 402--419. Springer, 2020 b

    2020

  40. [48]

    and Deng, J

    Teed, Z. and Deng, J. DROID-SLAM: Deep Visual SLAM for Monocular, Stereo, and RGB-D Cameras . Advances in neural information processing systems, 2021

    2021

  41. [49]

    Wang, J. Y. and Adelson, E. H. Representing moving images with layers. IEEE transactions on image processing, 3 0 (5): 0 625--638, 1994

    1994

  42. [50]

    Tracking everything everywhere all at once

    Wang, Q., Chang, Y.-Y., Cai, R., Li, Z., Hariharan, B., Holynski, A., and Snavely, N. Tracking everything everywhere all at once. In International Conference on Computer Vision, 2023 a

    2023

  43. [51]

    Deepvo: Towards end-to-end visual odometry with deep recurrent convolutional neural networks

    Wang, S., Clark, R., Wen, H., and Trigoni, N. Deepvo: Towards end-to-end visual odometry with deep recurrent convolutional neural networks. In 2017 IEEE International Conference on Robotics and Automation (ICRA), pp.\ 2043--2050. IEEE, 2017

    2017

  44. [52]

    Tartanvo: A generalizable learning-based vo

    Wang, W., Hu, Y., and Scherer, S. Tartanvo: A generalizable learning-based vo. arXiv preprint arXiv:2011.00359, 2020

    2011 arXiv

  45. [53]

    Neural video depth stabilizer

    Wang, Y., Shi, M., Li, J., Huang, Z., Cao, Z., Zhang, J., Xian, K., and Lin, G. Neural video depth stabilizer. arXiv preprint arXiv:2307.08695, 2023 b

    2023 arXiv

  46. [54]

    Dynamic-slam: Semantic monocular visual localization and mapping based on deep learning in dynamic environment

    Xiao, L., Wang, J., Qiu, X., Rong, Z., and Zou, X. Dynamic-slam: Semantic monocular visual localization and mapping based on deep learning in dynamic environment. Robotics and Autonomous Systems, 117: 0 1--16, 2019

    2019

  47. [55]

    Gmflow: Learning optical flow via global matching

    Xu, H., Zhang, J., Cai, J., Rezatofighi, H., and Tao, D. Gmflow: Learning optical flow via global matching. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 8121--8130, 2022

    2022

  48. [56]

    Self-supervised video object segmentation by motion grouping

    Yang, C., Lamdouar, H., Lu, E., Zisserman, A., and Xie, W. Self-supervised video object segmentation by motion grouping. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp.\ 7177--7188, 2021

    2021

  49. [57]

    v., Wang, R., and Cremers, D

    Yang, N., Stumberg, L. v., Wang, R., and Cremers, D. D3vo: Deep depth, deep pose and deep uncertainty for monocular visual odometry. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 1281--1292, 2020

    2020

  50. [58]

    Deflowslam: Self-supervised scene motion decomposition for dynamic dense slam

    Ye, W., Lan, X., Chen, S., Ming, Y., Yu, X., Li, J., Bao, H., Cui, Z., and Zhang, G. Deflowslam: Self-supervised scene motion decomposition for dynamic dense slam. 2022

    2022

  51. [59]

    PVO: Panoptic Visual Odometry

    Ye, W., Lan, X., Chen, S., Ming, Y., Yu, X., Bao, H., Cui, Z., and Zhang, G. PVO: Panoptic Visual Odometry . In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pp.\ 9579--9589, June 2023

    2023

  52. [60]

    Consistent depth maps recovery from a video sequence

    Zhang, G., Jia, J., Wong, T.-T., and Bao, H. Consistent depth maps recovery from a video sequence. IEEE Transactions on Pattern Analysis and Machine Intelligence, 31 0 (6): 0 974--988, 2009. doi:10.1109/TPAMI.2009.52

    2009 doi

  53. [61]

    Learning two-view correspondences and geometry using order-aware network

    Zhang, J., Sun, D., Luo, Z., Yao, A., Zhou, L., Shen, T., Chen, Y., Quan, L., and Liao, H. Learning two-view correspondences and geometry using order-aware network. In Proceedings of the IEEE/CVF international conference on computer vision, pp.\ 5845--5854, 2019

    2019

  54. [62]

    Particlesfm: Exploiting dense point trajectories for localizing moving cameras in the wild

    Zhao, W., Liu, S., Guo, H., Wang, W., and Liu, Y.-J. Particlesfm: Exploiting dense point trajectories for localizing moving cameras in the wild. In European conference on computer vision (ECCV), 2022

    2022

  55. [63]

    Motion-attentive transition for zero-shot video object segmentation

    Zhou, T., Wang, S., Zhou, Y., Yao, Y., Li, J., and Shao, L. Motion-attentive transition for zero-shot video object segmentation. In Proceedings of the 34th AAAI Conference on Artificial Intelligence (AAAI), pp.\ 13066--13073, 2020

    2020

  56. [64]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

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