Pith. sign in

REVIEW 4 major objections 4 minor 32 references

PickScan: Object discovery and reconstruction from handheld interactions

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

Pith's one-line read PickScan shows that picking up an object and moving it in front of an RGB-D camera is enough to discover, track, and reconstruct it as its own 3D model, with no training on object classes.

desk verdict A plausible class-agnostic interaction-based discovery pipeline whose headline reconstruction numbers are undercut by an apples-to-oranges comparison with Co-Fusion, but the core idea and false-positive reduction look real. read the letter →

arxiv 2411.11196 v1 pith:4FLVRXYT submitted 2024-11-17 cs.CV cs.AIcs.GRcs.LGcs.RO

classification cs.CVcs.AIcs.GRcs.LGcs.RO
keywords interaction-guidedobjectdiscoveryclass-agnostic3DreconstructioncompositionalsceneegocentricRGB-Dscanningmanipulationmovingmaskheuristictrackingunknown
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

PickScan claims that compositional 3D scene reconstruction, where each object gets its own model, can be reduced to finding one reliable mask per manipulated object. The paper's central idea is to use the act of picking an object up as the discovery signal: a displacement heuristic separates moving scene parts from the static background, and the crossing pattern of hand-to-scene versus hand-to-object distances marks the start and end of each interaction. This yields one high-quality mask per object, which a mask tracker spreads over the whole interaction and a neural reconstruction method turns into a 3D model. On a custom tabletop dataset, the pipeline reports 78.3% precision at 100% recall, a mean chamfer distance of 0.90 cm, and a 73% reduction in reconstruction error and 99% fewer false positives compared with the only comparable class-agnostic interaction-based baseline.

What carries the argument

The load-bearing mechanism is the moving-mask heuristic followed by a distance-crossing interaction detector. A pixel is labeled moving when its reprojection into the initial point cloud lies farther than a threshold $\rho_\mathrm{moving}$ from every initial-scene point; subtracting the estimated arm mask from the moving mask leaves candidate object blobs, and the blob whose points lie closest to the hand is kept as the candidate object mask. Interaction boundaries come from two distance trajectories: the distance between the hand point cloud and the initial scene, and the distance between the hand point cloud and the candidate object. During a pick-up-and-lay-down maneuver the first rises above the second at the start and falls back below it at the end, so those crossings segment individual interactions. A median filter and a minimum-duration rule remove noise, the longest period of stable cross-frame intersection-over-union supplies a 'best frame' whose mask seeds 2D tracking, and a containment-based non-maximum suppression step removes duplicate detections after tracking.

What would settle it

Scan a tabletop scene in which a second rigid object moves independently, for example a ball rolls across the table, while the user picks up the target object, and count the pipeline's false-positive object detections: if they rise well above the reported one per scan, the discovery mechanism depends on the scene being otherwise static, and the central claim is correspondingly limited.

Watch

Extended reading notes

Core claim

The central claim is that the act of picking up an object carries enough information to discover and segment that object without any object-class prior. The paper's contribution is an interaction detector that reads the crossing pattern of two distances in an egocentric RGB-D stream: the hand-to-initial-scene distance crosses above the hand-to-object distance when the object is lifted and falls below it when the object is laid down, marking the start and end of one object interaction. From within each interaction, the pipeline selects the frame whose candidate mask is most stable across neighboring frames, tracks that mask through the entire interaction with a 2D mask tracker, and feeds the tracked masks to an unknown-object reconstruction method that outputs a 3D model and per-frame pose. On the paper's three-scene tabletop dataset, this yields 100% recall with 78.3% precision, a mean chamfer distance of 0.90 cm, a 73% reduction in reconstruction error, and 99% fewer false positives compared with the only comparable class-agnostic interaction-based baseline.

Load-bearing premise

The whole discovery step rests on the assumption that in the egocentric video the only moving things are the user's arm and the object being picked up, and that the object moves far enough from its original position to be flagged as moving; if another thing moves or the object is only slightly nudged, the pipeline cannot isolate it.

Editorial extensions

If this is right

  • Users can scan any rigid, graspable object in a scene with a handheld RGB-D camera and receive one complete 3D model per picked-up object, bypassing the training-class limits of segmentation-based discovery.
  • Interaction-based discovery resolves object boundaries that static segmentation cannot, such as adjacent bricks with identical texture but different shapes, because motion separates the picked-up object from its support.
  • Per-pixel displacement detection replaces superpixel-center clustering, yielding substantially cleaner masks and a 73% lower mean chamfer distance than the prior class-agnostic baseline.
  • The pipeline's modularity means any improvement in 2D mask tracking or unknown-object reconstruction translates directly into better end-to-end compositional reconstruction without changing the discovery stage.

Reading between the lines

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

  • The crossing-signal detector is a generic signature of grasp-and-move actions, so the same idea could drive object discovery in robot demonstration learning or augmented-reality object editing, where a hand-object distance staying small while the hand-to-scene distance grows marks the manipulation.
  • An untested stress case is a scene with a second, independent mover (a person walking by or a rolling object); because the moving-mask heuristic assumes only the arm and held object move, such a scene would likely inflate false positives and reveal the boundary of the static-background assumption.
  • Because the discovery stage only needs one good mask per object, running the interaction detector online could turn the pipeline into an interactive scanner that tells the user when an object has been seen from enough viewpoints; the paper does not pursue real-time feedback.
  • The reported performance is measured on a three-scene tabletop dataset with five known foreground objects, so the class-agnostic claim would be strengthened by tests on unseen objects, varied supports such as shelves or floors, and lower-quality depth sensors; these are direct extensions the paper leaves open.
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 PickScan, an interaction-guided and class-agnostic pipeline for discovering and reconstructing manipulated objects from egocentric RGB-D/IMU streams. The method builds an initial static point cloud with gradSLAM, computes per-frame 'moving masks' by reprojecting current depth into that cloud and thresholding at a distance rho_moving, removes an arm mask, selects blobs near the hand, detects pick-up/lay-down interactions from crossings of the hand-initial and hand-object distance trajectories, and chooses one stable mask per interaction. The mask is tracked with XMem and the object is reconstructed with BundleSDF. On a custom dataset of three tabletop scans (nine manipulated objects), the paper reports a mean chamfer distance of 0.90 cm (versus Co-Fusion's 3.33 cm), 78.3% precision at 100% recall, and an average of 1.0 false positive per scan. A segmentation-based baseline (SAM*) is also compared.

Significance. The core idea is attractive: by exploiting the user's physical interaction with an object, PickScan avoids class-specific appearance priors and can separate objects that static segmentation merges, as illustrated by the toy-brick example in Fig. 8. The system is modular, reuses established components (XMem, BundleSDF, gradSLAM), and the authors share code. The paper is also candid about its limitations, including false positives, mask-tracking failures, and reconstruction gaps attributed to BundleSDF. If the quantitative claims are confirmed by a controlled evaluation, the work would be a useful step toward practical compositional scanning. However, the current evidence is preliminary: the dataset is small, threshold values are not reported, and the Co-Fusion chamfer comparison is confounded by the reconstruction backend. The precision and false-positive metrics are less affected by the backend confound but are based on only nine positive examples.

major comments (4)
  1. [Sec. IV-D, Table I] The comparison with Co-Fusion is confounded by the reconstruction backend. PickScan uses BundleSDF (Sec. III, pipeline overview) and the SAM* baseline explicitly reuses XMem+BundleSDF (Sec. IV-E), but the paper never specifies how Co-Fusion's reconstructions were obtained. If Co-Fusion was run as a complete system with its own superpixel-based fusion, the reported 73% chamfer-distance reduction cannot be attributed to the proposed interaction-based masks; BundleSDF's tracking and fusion could account for much of the gap. Please add a controlled experiment in which the same reconstruction backend (e.g., XMem+BundleSDF) is fed with masks produced by PickScan and by Co-Fusion, and report the resulting chamfer distances. Without this, the central claim of outperforming Co-Fusion in every metric is not established.
  2. [Sections III-A through III-E and Table I] The method has five free thresholds that are described only as 'tunable': rho_moving (III-A), the blob area threshold (III-B), the interaction-duration threshold (III-C), tau_iou (III-D), and tau_nms (III-E). Their values are not reported, and the 'topmost 10%' hand-mask fraction is stated without supporting evidence. The Table I numbers (precision, false positives, chamfer distance) are directly controlled by these parameters. Please report the exact values used for all thresholds and include a sensitivity analysis that varies each threshold over a plausible range and shows the effect on precision, recall, and chamfer distance. Without this, the headline results are not reproducible.
  3. [Sec. IV-A, Table I] The evaluation uses only three scans and nine manipulated objects drawn from a set of five foreground objects. There are no error bars, no per-object results, and no significance test; the per-scene precision values (100%, 75%, 60%) show considerable variance. A claim of 100% recall based on nine positive examples is fragile. Please provide per-object or per-scene results with error bars and, ideally, evaluate on a larger and more diverse set of scenes and objects to support the general conclusions.
  4. [Sec. III-A and III-C] The core discovery heuristic assumes that the only moving entities in the egocentric stream are the user's arm and the manipulated object, and that a manipulated object's points are displaced farther than rho_moving from the initial point cloud. The paper does not test what happens when this assumption is violated, for example when a second person or another dynamic object enters the view, or when an object is slid by only a few centimeters. Since this assumption is the foundation of the object-discovery mechanism, the paper should either explicitly scope the claimed generality to this setting or include a stress-test experiment with such violations. This is a correctness-risk concern rather than an objection to the overall idea.
minor comments (4)
  1. [Introduction, contributions paragraph] The phrase 'beginnings and ends of individual user-object detections' appears to be a typo; it should read 'interactions' rather than 'detections'.
  2. [Introduction and References] Co-Fusion is cited as [10] in the Introduction, but reference [10] is the semantic visual SLAM survey by Chen et al.; the correct citation for Co-Fusion is [5] (Rünz and de Agapito).
  3. [Sec. III-A] The reprojection equation is difficult to parse because the homogeneous-coordinate notation and the definitions of p_t(u,v), p_0(u,v), and T_{0,t} are not written cleanly. Please rewrite the equation with explicit 4-vector notation so that the displacement heuristic is reproducible.
  4. [Sec. IV-A] The dataset description does not specify how many times each of the five foreground objects appears across the three scenes; please add a per-scene object list so that the per-scene metrics in Table I can be interpreted.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the pipeline is an empirical heuristic system evaluated against externally scanned ground-truth models, and the cited prior works are independent components rather than load-bearing self-justifications.

full rationale

The paper does not claim a first-principles derivation of its reconstructions. Its core object-discovery mechanism is a displacement heuristic (Section III-A): pixels are labeled moving if they reproject farther than a tunable radius rho_moving from an initial point cloud, and interactions are detected by the crossing pattern of hand-initial versus hand-object distances (Section III-C). These are operational heuristics, not parameters fitted to the target chamfer distance or recall values. The quantitative claims are measured against high-fidelity ground-truth models obtained with an external static scanner ([30]), so the evaluation is self-contained and externally anchored. The thresholds rho_moving, tau_iou, and tau_nms are stated as tunable, and there is no evidence they were back-solved from the reported 0.90 cm chamfer distance or 78.3% precision. The comparison to Co-Fusion conflates object-discovery quality with the stronger BundleSDF reconstruction backend, but that is an experimental confound and a fairness concern, not a circularity: the Co-Fusion numbers are external measurements, not derived from the paper's own equations. The paper does cite the current authors' own work, notably gradSLAM [29] for point-cloud fusion and ConceptFusion [13] in related work, but neither is load-bearing in a way that forces the paper's conclusions: gradSLAM is an independently released SLAM pipeline used as a component, and ConceptFusion is only contextual. No uniqueness theorem is imported from the authors' prior work, and no 'prediction' is equivalent by construction to an input. Under the stated rules, the central claim retains independent empirical content, so the circularity score is 0.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim rests on several heuristic thresholds and domain assumptions. No new physical entities are introduced. The thresholds are named but not specified, which is the main burden on reproducibility and on the strength of the reported metrics.

free parameters (6)
  • rho_moving
    Displacement threshold in Section III-A for classifying pixels as moving. Tunable, but no default or selected value is reported.
  • blob area threshold
    Minimum contiguous pixel area for candidate object blobs in Section III-B. No value reported.
  • interaction duration threshold
    Minimum duration for an interaction in Section III-C. No value reported.
  • tau_iou
    Cross-frame IoU threshold for stable periods in Section III-D. No value reported.
  • tau_nms
    Mask containment threshold for duplicate removal in Section III-E. No value reported.
  • top 10% hand mask fraction = 0.10
    Fixed fraction of the arm mask's vertical span used as the hand mask in Section III-B. Chosen by hand without sensitivity analysis.
assumptions (4)
  • domain assumption In the egocentric stream, the only moving objects are the user's arm and a possibly manipulated object.
    Section III-A states this directly; false when other dynamic scene content exists, such as another person or moving furniture, which would enter the moving mask.
  • domain assumption A manipulated object's points are sufficiently far from the initial point cloud to be detected by threshold rho_moving.
    Section III-A defines moving pixels via distance to the initial point cloud; objects that remain near their original support, such as small local sliding, may not be detected.
  • domain assumption The static scanning phase precedes all manipulation and the arm mask becomes non-empty exactly at the end of the static phase.
    Section III-A uses pretrained arm segmentation to determine when the static scan ends; failure of the arm detector shifts the initial point cloud.
  • ad hoc to paper User-object interactions produce a characteristic crossing of hand-initial distance above hand-object distance followed by a crossing below.
    Section III-C postulates this temporal signature for picking up and laying down; not derived from a model of grasping, so it may miss interactions that do not follow the assumed pattern.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PickScan: Object discovery and reconstruction from handheld interactions." pith.science (2026). https://pith.science/paper/4FLVRXYT

@misc{pith2026241111196,
  author       = {Pith},
  title        = {Pith review of: PickScan: Object discovery and reconstruction from handheld interactions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4FLVRXYT}},
  note         = {Machine review of arXiv:2411.11196}
}
read the original abstract

Reconstructing compositional 3D representations of scenes, where each object is represented with its own 3D model, is a highly desirable capability in robotics and augmented reality. However, most existing methods rely heavily on strong appearance priors for object discovery, therefore only working on those classes of objects on which the method has been trained, or do not allow for object manipulation, which is necessary to scan objects fully and to guide object discovery in challenging scenarios. We address these limitations with a novel interaction-guided and class-agnostic method based on object displacements that allows a user to move around a scene with an RGB-D camera, hold up objects, and finally outputs one 3D model per held-up object. Our main contribution to this end is a novel approach to detecting user-object interactions and extracting the masks of manipulated objects. On a custom-captured dataset, our pipeline discovers manipulated objects with 78.3% precision at 100% recall and reconstructs them with a mean chamfer distance of 0.90cm. Compared to Co-Fusion, the only comparable interaction-based and class-agnostic baseline, this corresponds to a reduction in chamfer distance of 73% while detecting 99% fewer false positives.

Figures

Figures reproduced from arXiv: 2411.11196 by the authors.

Figure 1
Figure 1. We present PickScan, an interaction-guided and class-agnostic pipeline for compositional scene reconstruction. Our method lets a user pick up and move around objects, and outputs the object masks, 3D model and per-frame poses of each manipulated object. Abstract— Reconstructing compositional 3D representations of scenes, where each object is represented with its own 3D model, is a highly desirable capability in robo… view at source ↗
Figure 2
Figure 2. Example illustrating the benefit of user-object interactions [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Overview of our pipeline and breakdown of its two phases: [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Interaction detection visualized: User-object interactions are [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Scenes 1 to 3, from left to right, of our dataset. The images [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Output of our pipeline on an example scene from our [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]
Figure 7
Figure 7. Figure 7: Qualitative comparison between the object masks provided [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 28 canonical work pages

  1. [1]

    Kimera: an open- source library for real-time metric-semantic localization and mapping,

    A. Rosinol, M. Abate, Y . Chang, and L. Carlone, “Kimera: an open- source library for real-time metric-semantic localization and mapping,” 2020 IEEE International Conference on Robotics and Automation (ICRA), pp. 1689–1696, 2019. 1, 2

  2. [2]

    Se- manticfusion: Dense 3d semantic mapping with convolutional neural networks,

    J. McCormac, A. Handa, A. J. Davison, and S. Leutenegger, “Se- manticfusion: Dense 3d semantic mapping with convolutional neural networks,” 2017 IEEE International Conference on Robotics and Automation (ICRA) , pp. 4628–4635, 2016. 1, 2

  3. [3]

    Fusion++: V olumetric object-level slam,

    J. McCormac, R. Clark, M. Bloesch, A. J. Davison, and S. Leuteneg- ger, “Fusion++: V olumetric object-level slam,” 2018 International Conference on 3D Vision (3DV) , pp. 32–41, 2018. 1, 2

  4. [4]

    Real-time and scalable incre- mental segmentation on dense slam,

    K. Tateno, F. Tombari, and N. Navab, “Real-time and scalable incre- mental segmentation on dense slam,” in 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2015, pp. 4465–

  5. [5]

    Co-fusion: Real-time segmentation, tracking and fusion of multiple objects,

    M. R ¨unz and L. de Agapito, “Co-fusion: Real-time segmentation, tracking and fusion of multiple objects,” 2017 IEEE International Conference on Robotics and Automation (ICRA) , pp. 4471–4478,

  6. [6]

    Dynaslam ii: Tightly-coupled multi-object tracking and slam,

    B. Besc ´os, C. Campos, J. D. Tard ´os, and J. Neira, “Dynaslam ii: Tightly-coupled multi-object tracking and slam,” IEEE Robotics and Automation Letters, vol. 6, pp. 5191–5198, 2020. 1, 2

  7. [7]

    Mid-fusion: Octree-based object-level multi-instance dynamic slam,

    B. Xu, W. Li, D. Tzoumanikas, M. Bloesch, A. J. Davison, and S. Leutenegger, “Mid-fusion: Octree-based object-level multi-instance dynamic slam,” 2019 International Conference on Robotics and Au- tomation (ICRA) , pp. 5231–5237, 2018. 1, 2

  8. [8]

    Maskfusion: Real-time recognition, tracking and reconstruction of multiple moving objects,

    M. R ¨unz and L. de Agapito, “Maskfusion: Real-time recognition, tracking and reconstruction of multiple moving objects,” 2018 IEEE International Symposium on Mixed and Augmented Reality (ISMAR) , pp. 10–20, 2018. 1, 2

Show all 32 references
  1. [9]

    Segment anything,

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Lo, P. Doll ´ar, and R. Gir- shick, “Segment anything,” arXiv:2304.02643, 2023. 2, 6

  2. [10]

    Semantic visual simultaneous localization and mapping: A survey,

    K. Chen, J. Zhang, J. Liu, Q. Tong, R. Liu, and S. Chen, “Semantic visual simultaneous localization and mapping: A survey,” ArXiv, vol. abs/2209.06428, 2022. 2

  3. [11]

    Xmem: Long-term video object segmentation with an atkinson-shiffrin memory model,

    H. K. Cheng and A. G. Schwing, “Xmem: Long-term video object segmentation with an atkinson-shiffrin memory model,” in European Conference on Computer Vision , 2022. 2, 3, 4

  4. [12]

    Bundlesdf: Neural 6-dof tracking and 3d reconstruction of unknown objects,

    B. Wen, J. Tremblay, V . Blukis, S. Tyree, T. Muller, A. Evans, D. Fox, J. Kautz, and S. Birchfield, “Bundlesdf: Neural 6-dof tracking and 3d reconstruction of unknown objects,” 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 606–617, 2023. 2, 3

  5. [13]

    Conceptfusion: Open-set multimodal 3d mapping,

    K. M. Jatavallabhula, A. Kuwajerwala, Q. Gu, M. Omama, T. Chen, S. Li, G. Iyer, S. Saryazdi, N. V . Keetha, A. K. Tewari, J. B. Tenenbaum, C. M. de Melo, M. Krishna, L. Paull, F. Shkurti, and A. Torralba, “Conceptfusion: Open-set multimodal 3d mapping,” ArXiv, vol. abs/2302.07...

  6. [14]

    Nesf: Neural semantic fields for generalizable semantic segmentation of 3d scenes,

    S. V ora, N. Radwan, K. Greff, H. Meyer, K. Genova, M. S. M. Sajjadi, E. Pot, A. Tagliasacchi, and D. Duckworth, “Nesf: Neural semantic fields for generalizable semantic segmentation of 3d scenes,” ArXiv, vol. abs/2111.13260, 2021. 2

  7. [15]

    Reconstructing hand-held objects from monocular video,

    D. Huang, X. Ji, X. H. He, J. Sun, T. He, Q. Shuai, W. Ouyang, and X. Zhou, “Reconstructing hand-held objects from monocular video,” SIGGRAPH Asia 2022 Conference Papers , 2022. 2

  8. [16]

    In-hand 3d object scanning from an rgb sequence,

    S. Hampali, T. Hodan, L. Tran, L. Ma, C. Keskin, and V . Lepetit, “In-hand 3d object scanning from an rgb sequence,” 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pp. 17 079–17 088, 2022. 2

  9. [17]

    Demograsp: Few-shot learning for robotic grasping with human demonstration,

    P. Wang, F. Manhardt, L. Minciullo, L. Garattoni, S. Meie, N. Navab, and B. Busam, “Demograsp: Few-shot learning for robotic grasping with human demonstration,” 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pp. 5733–5740, 2021. 2

  10. [18]

    3d object reconstruction from hand-object interactions,

    D. Tzionas and J. Gall, “3d object reconstruction from hand-object interactions,” 2015 IEEE International Conference on Computer Vision (ICCV), pp. 729–737, 2015. 2

  11. [19]

    On the analysis of accumulative difference pictures from image sequences of real world scenes,

    R. Jain and H.-H. Nagel, “On the analysis of accumulative difference pictures from image sequences of real world scenes,” IEEE Trans- actions on Pattern Analysis and Machine Intelligence , vol. PAMI-1, no. 2, pp. 206–214, 1979. 2

  12. [20]

    Statistical background subtraction for a mobile observer,

    E. Hayman and J.-O. Eklundh, “Statistical background subtraction for a mobile observer,” in Proceedings Ninth IEEE International Conference on Computer Vision , 2003, pp. 67–74 vol.1. 2

  13. [21]

    Sparse subspace clustering: Algorithm, theory, and applications,

    E. Elhamifar and R. Vidal, “Sparse subspace clustering: Algorithm, theory, and applications,” IEEE transactions on pattern analysis and machine intelligence, 2012. 2

  14. [22]

    Object segmentation by long term analysis of point trajectories,

    T. Brox and J. Malik, “Object segmentation by long term analysis of point trajectories,” in European Conference on Computer Vision, 2010. 2

  15. [23]

    Flownet3d: Learning scene flow in 3d point clouds,

    X. Liu, C. Qi, and L. J. Guibas, “Flownet3d: Learning scene flow in 3d point clouds,” 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pp. 529–537, 2018. 2

  16. [24]

    Multimotion visual odometry (mvo): Simultaneous estimation of camera and third-party motions,

    K. M. Judd, J. D. Gammell, and P. Newman, “Multimotion visual odometry (mvo): Simultaneous estimation of camera and third-party motions,” 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pp. 3949–3956, 2018. 2

  17. [25]

    D2nerf: Self-supervised decoupling of dynamic and static objects from a monocular video,

    T. W. Wu, F. Zhong, A. Tagliasacchi, F. Cole, and C. Oztireli, “D2nerf: Self-supervised decoupling of dynamic and static objects from a monocular video,” in Advances in Neural Information Processing Systems, 2022. 2

  18. [26]

    Neuraldiff: Segmenting 3d objects that move in egocentric videos,

    V . Tschernezki, D. Larlus, and A. Vedaldi, “Neuraldiff: Segmenting 3d objects that move in egocentric videos,”2021 International Conference on 3D Vision (3DV) , pp. 910–919, 2021. 2

  19. [27]

    D- nerf: Neural radiance fields for dynamic scenes,

    A. Pumarola, E. Corona, G. Pons-Moll, and F. Moreno-Noguer, “D- nerf: Neural radiance fields for dynamic scenes,” 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pp. 10 313–10 322, 2020. 3

  20. [28]

    Fine-grained egocentric hand- object segmentation: Dataset, model, and applications,

    L. Zhang, S. Zhou, S. Stent, and J. Shi, “Fine-grained egocentric hand- object segmentation: Dataset, model, and applications,” in European Conference on Computer Vision , 2022. 3

  21. [29]

    ∇slam: Dense slam meets automatic differentiation,

    K. M. Jatavallabhula, G. Iyer, and L. Paull, “ ∇slam: Dense slam meets automatic differentiation,” 2020 IEEE International Conference on Robotics and Automation (ICRA) , pp. 2130–2137, 2019. 3

  22. [30]

    Einscan v2 3d scanner webpage,

    “Einscan v2 3d scanner webpage,” https://www.einscan.com/ einscan-se/, accessed: 2024-01-09. 4

  23. [31]

    Camera icon,

    “Camera icon,” https://www.flaticon.com/free-icon/video-camera 1159798, accessed: 2024-03-15. 7

  24. [32]

    Hand icon,

    “Hand icon,” https://www.flaticon.com/free-icon/grab 5698645, ac- cessed: 2024-03-15. 7

Pith tools

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