Pith. sign in

REVIEW 4 major objections 5 minor 54 references

ToF-Splatting: Dense SLAM using Sparse Time-of-Flight Depth and Multi-Frame Integration

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read ToF-Splatting claims that 64-point 8×8 ToF measurements, integrated across keyframes with monocular cues, are enough to run accurate 3D Gaussian Splatting SLAM.

desk verdict First 3DGS-SLAM that handles 64-point ToF depth; strong ZJUL5 gains but missing ablations leave which component earns them unclear. read the letter →

arxiv 2504.16545 v1 pith:AXIAAMZK submitted 2025-04-23 cs.CV

Andrea Conti , Matteo Poggi , Valerio Cambareri , Martin R. Oswald , Stefano Mattoccia This is my paper · ORCID
classification cs.CV
keywords time-of-flightsparsedepthSLAM3DGaussianSplattingdensemappingmulti-frameintegrationcompletioncameratracking
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

This paper claims that very sparse time-of-flight (ToF) depth, as little as the 64 points from an $8\times 8$ sensor, is enough to run an accurate dense SLAM system if the sparse measurements are integrated across multiple frames and combined with monocular cues. The authors propose ToF-Splatting, the first 3D Gaussian Splatting-based SLAM pipeline built for such sparse ToF input. A multi-frame integration module produces dense metric depth maps that seed new Gaussians and supervise both tracking and mapping. On the real-world ZJUL5 dataset, the system reports an absolute trajectory error of 0.051 m versus 0.107 m for the prior ToF-SLAM baseline, and a higher mapping F-score. If correct, this shows that the extreme sparsity and noise of low-power ToF sensors need not block their use in mobile and AR/VR SLAM.

What carries the argument

The load-bearing component is the multi-frame integration module, an extension of the Depth on Demand framework. Given a target frame, it iteratively refines a dense depth prediction using a set of source keyframes selected for sufficient parallax, their relative poses, the current sparse ToF measurements, and the RGB image paired with a normalized monocular depth prior from Depth Anything v2. The output dense depth map $D_k$ serves three purposes: it seeds new Gaussians during mapping, it supervises the geometric term of the tracking loss, and it supervises the mapping optimization. Its outlier handling step produces a prior depth prediction without the ToF points, discards measurements above a quantile of the $\ell^1$ error against that prior, and repredicts with the cleaned sparse depth.

What would settle it

Run ToF-Splatting on a real sparse-ToF sequence with ground-truth depth and poses, disable the outlier filter, and check whether the filter removes points that lie within sensor noise of the ground truth; if such valid points are dropped, the filter is enforcing the network's prior rather than removing sensor outliers, and tracking error should track the network's depth error rather than the ToF noise level.

Watch

Extended reading notes

Core claim

The central discovery is that replacing single-frame depth completion with multi-frame integration makes a 3DGS-based SLAM system usable with sensors that return only 64 depth points per frame. The multi-frame integration module, an extension of Depth on Demand, combines the sparse ToF measurements with multiple keyframe views, the relative poses from tracking, and monocular depth priors from Depth Anything v2 to predict dense metric depth maps. These maps provide the geometric supervision that 3DGS-based SLAM needs; the system also filters ToF outliers by comparing each measurement against the network's prediction without ToF input. ToF-Splatting reports state-of-the-art tracking (ATE 0.051 m on ZJUL5) and mapping (F-score 0.664) among methods using sparse ToF, and on TUM RGB-D at 0.04% depth density it approaches the accuracy of a baseline using 100% dense depth.

Load-bearing premise

The system's tracking and mapping supervision rests on dense depth maps produced by a network retrained on ScanNet; if that network's depth prior is wrong for an unseen scene, errors in those maps dominate the optimization, and the outlier filter, which checks ToF points against the same prior, can throw away valid measurements.

Editorial extensions

If this is right

  • Low-power ToF sensors with tens of depth points become viable for dense SLAM in mobile and AR/VR devices, removing the need for higher-resolution, more power-hungry depth sensors.
  • 3DGS-based SLAM systems, previously thought to require dense and accurate depth, can be bootstrapped from very sparse measurements when multi-view geometry is integrated across a keyframe buffer.
  • At 0.04% depth density on TUM RGB-D, the method approaches the tracking accuracy of a baseline that receives 100% dense depth, suggesting sparse active depth plus geometry can nearly substitute for dense RGB-D input indoors.
  • The reported runtime of about 1.5 frames per second shows the pipeline is not real-time as published, so deployment depends on faster 3DGS-based SLAM backends.

Reading between the lines

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

  • Because the outlier filter compares each ToF measurement against the network's own no-ToF depth prediction, a systematically biased depth prior would silently discard valid measurements; the method's robustness ceiling is therefore set by how well the ScanNet-trained network transfers to the deployment scene.
  • The reported ZJUL5 and Replica results measure generalization to real and synthetic data, but the training sparsities stop at 0.03% and the test densities start at 0.02%, so a wider sweep of densities and sensor noise profiles would clarify where the gains saturate.
  • The multi-frame integration module could be evaluated in isolation with ground-truth poses and ToF measurements while varying the monocular prior, separating the contribution of multi-view geometry from that of the monocular cue.
  • The same keyframe-buffer integration idea transfers to NeRF-based SLAM or to depth completion over posed video, since the module only needs posed views and sparse depth as input.
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 / 5 minor

Summary. The paper proposes ToF-Splatting, a 3D Gaussian Splatting (3DGS) based dense SLAM system designed for extremely sparse Time-of-Flight depth input (e.g., 8x8 = 64 points per frame). The pipeline combines three modules: a tracking frontend that optimizes poses with photometric and geometric losses, a mapping backend that seeds and optimizes 3D Gaussians, and a multi-frame integration module that extends the authors' prior Depth on Demand (DoD) framework to fuse sparse ToF depth, multi-view geometry from a keyframe buffer, and monocular cues from Depth Anything V2. The integration module also includes an outlier filtering step based on the network's own no-ToF prediction. Experiments on the real-world ZJUL5 dataset report lower ATE than ToF-SLAM (0.051 m vs. 0.107 m) and higher mapping F-score (0.664 vs. 0.604), with additional simulated experiments on TUM RGB-D and Replica, runtime measurements, and an explicit limitations paragraph.

Significance. If the reported results hold up, ToF-Splatting is a meaningful advance: it is the first 3DGS-based SLAM system specifically adapted to sparse, noisy ToF sensors, and the ZJUL5 experiments demonstrate that a multi-frame integration strategy can support dense mapping from very few depth points. The paper does several things well: it evaluates on real sparse-ToF data rather than only simulations, it ablates the contribution of monocular versus multi-view cues within the integration module, it includes robustness studies for depth sparsity, noise, and temporal sparsity, and it reports runtime and limitations transparently. However, the central SOTA claim is not yet fully supported, because the experiments do not isolate the multi-frame integration module from the pretrained components (DoD, Depth Anything V2) that feed it, and the outlier filter is unablated and has a circular flavor. These are correctness-risk concerns rather than internal contradictions, and they are addressable with additional controlled experiments.

major comments (4)
  1. [§3.2, Outlier Handling] The outlier filter is load-bearing for the ZJUL5 results, since real ToF points are described as very noisy and outlier-prone, yet it is not ablated and its design raises a circularity concern. The filter computes a no-ToF depth prediction with the same DoD network, discards ToF measurements whose L1 error against that prediction exceeds quantile q=0.75, and then runs the same network again using the filtered points. If the no-ToF prediction is biased, this procedure removes valid measurements and enforces the network's own hallucinated depth instead of correcting it. The paper should report an ablation of the filter (filter on/off), the fraction of points discarded per sequence, and ideally compare the filter against an independent geometric consistency check or against filtering with a different reference depth.
  2. [§4.2, Ablation Studies and Tables 1–3] The central claim that 'multi-frame integration' is responsible for the large gains over ToF-SLAM is not directly supported by the experiments. Table 3 ablates monocular and multi-view cues inside the DoD-based module, but there is no ablation that replaces the module itself -- for example, using raw sparse depth, DELTAR-densified depth, or single-frame DoD depth for seeding and supervision while keeping the rest of the pipeline fixed. Without such a controlled comparison, the Table 1–2 improvements could be attributed to the retrained DoD network (the authors' own prior work) and the Depth Anything V2 priors rather than to the multi-view integration, keyframe handling, or the 3DGS backend. This is the key experiment needed to support the stated state-of-the-art claim.
  3. [Tables 1 and 2] The quantitative evaluation on ZJUL5 is reported without error bars or multiple runs, and the dataset contains only seven real sequences. Given that several hyperparameters (q=0.75, νth=0.1, σ=0.98, λtrack, λmap, N=4, b=15 cm) appear manually selected and are not varied in a sensitivity analysis, it is difficult to assess whether the 2x ATE margin over ToF-SLAM is robust. The paper should state the hyperparameter selection protocol (e.g., validation split or sensitivity plots) and report variance across repeats or at least per-sequence stability. The claim of state-of-the-art on 'reference datasets' also rests on a single real dataset; the TUM and Replica results are simulated and are not compared against the same baselines used on ZJUL5.
  4. [§4.1, Table 2] The mapping comparison with ToF-SLAM is partially ambiguous because, as the paper states, 'for each scene, we collect the predicted pose of each frame and render depth and color from 3DGS' for the ToF-Splatting meshes, whereas the ToF-SLAM numbers are presumably produced by its own pipeline. It would strengthen the comparison to confirm that the same reconstruction protocol (TSDF voxel size, truncation, marching cubes parameters) is applied to both methods, since mesh-based F-score is sensitive to these choices. At minimum, the paper should state explicitly whether the protocol for all baselines is identical and whether any method's reported numbers are taken from the original paper rather than recomputed under the same protocol.
minor comments (5)
  1. [Abstract and §4.1, Table 1] The text says 'ZJUL5 dataset' has '8 sequences available' in the Table 1 caption, while Section 4 states seven indoor scene recordings and Table 2 lists seven named scenes. Please correct this inconsistency.
  2. [§3.4, Eq. (10)] The notation in the isotropy loss is confusing: the text says diag(·) extracts the diagonal values and then refers to the average of the resulting vector, but the equation writes ||diag(Sj) - diag(Sj)·1_{3x1}||_1, which suggests diag(Sj) is a scalar. Please make the distinction between the diagonal vector and its mean explicit.
  3. [§3.4, Initialization] The sentence 'we prove this approach effective' overstates what is shown; the keyframe selection and seeding strategy are not ablated in the experiments. Please rephrase to 'we show empirically' or add an ablation, and in the meantime the claim should be softened.
  4. [§4.2, Table 4] In the TUM comparison, ToF-Splatting's 0.02%/0.04% simulated density results are compared with RGB-only methods and with a 100% density RGB-D MonoGS baseline. Please clarify that the RGB-D 100% row is an upper-bound reference and not a sparse-input competitor, and describe how the sparse points are sampled (e.g., random versus grid) since this affects the comparison.
  5. [Supplementary, §9 Temporal Sparsity] The temporal sparsity experiment provides first 50 frames with ToF depth to establish scale, but the figure and text do not state how the scale is maintained afterward for frames without ToF measurements. A brief explanation of how the multi-frame module obtains metric scale in those frames would help the reader interpret the graceful degradation shown in Figure 10.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the DoD-based depth network is retrained on external ScanNet and evaluated on external benchmarks; the pose-depth dependence is an iterative refinement loop, not a fitted-input prediction.

full rationale

The claimed derivation is not circular. The dense supervision D_k is produced by a multi-frame integration network that, although based on the authors' prior DoD work [3], is explicitly retrained from scratch on ScanNetv2 with altered training data and added monocular cues (Sec. 3.2 and Sec. 10), so it is not fitted to the ZJUL5/TUM/Replica test outputs that support the SOTA claim. The apparent pose-depth loop is a two-stage refinement: tracking first minimizes only the photometric loss to obtain a pose (Eqs. 4-5), that pose is then used to generate D_k, and a fixed D_k supervises further pose optimization; no parameter is fit to the final trajectory or map. The outlier filter that compares ToF points against the network's no-ToF prediction is self-referential and could propagate prior bias, but it does not make D_k equal to the no-ToF prior by construction, nor is it a fitted test-time parameter renamed as a prediction. Self-citations to DoD [3] and to the keyframe-based multi-view depth method [4] supply the architecture, but the component is retrained and independently evaluated, and the reported SOTA numbers are obtained against external baselines on external datasets. Missing ablations (e.g., replacing the retrained DoD module with DELTAR densification) are an evidence gap and a correctness risk, not circularity.

Assumptions & free parameters 9 free parameters · 6 assumptions · 0 invented entities

The central claims rest on a stack of pretrained or retrained components and hand-chosen hyperparameters. The most sensitive choices are the ScanNet retraining distribution, the q=0.75 outlier cutoff, and the assumption that DoD's no-ToF prediction is a valid reference for filtering. No new physical or mathematical entities are introduced.

free parameters (9)
  • outlier quantile q = 0.75
    Discards ToF points whose error against the no-ToF DoD depth exceeds the 0.75 quantile (Sec. 3.2); chosen because outliers are frequent in ZJUL5, so it steers the depth supervision.
  • baseline distance b = 15 cm
    Source keyframes are selected to have relative pose baseline near 15 cm to ensure parallax (Sec. 3.2).
  • number of source views N = 4
    Multi-frame integration uses N=4 keyframes per prediction (Sec. 3.2 and Implementation Details).
  • keyframe threshold nu_th = 0.1
    A frame becomes a keyframe when its opacity-based novelty factor exceeds 0.1 (Sec. 3.3).
  • opacity uncertainty threshold sigma = 0.98
    Rendered opacity below 0.98 marks uncertain pixels for keyframe selection and Gaussian seeding (Sec. 3.3).
  • tracking steps eta_rgb, eta_rgbd and lambda_track = 30, 70, 0.9
    Tracking runs 30 photometric-only steps then 70 RGB-D steps, with lambda_track=0.9 between photometric and depth loss (Sec. 3.3).
  • mapping weights and steps = lambda_map=60, lambda_visual=0.20, lambda_normals=0.01, lambda_iso=1.0, eta_M=60
    Mapping loss weights and optimization step count are set by hand (Sec. 3.4).
  • Gaussian seeding random downsampling factor = not specified
    Newly seeded Gaussians are randomly downsampled by a constant factor (Sec. 3.4); the factor is not reported, hampering exact replication.
  • DoD retraining sparse depth density range = 0% to 0.03%
    The integration network is retrained on ScanNet with random target sparsity up to 0.03%, matching the extreme sparsity of the test regime (Supp. Sec. 10).
assumptions (6)
  • standard math 3D Gaussian Splatting projection and rendering equations (Eqs. 1-3) are valid for the mapping and tracking optimization.
    The method relies on the differentiable splatting formulation from Kerbl et al. [13].
  • domain assumption DoD, retrained on ScanNet with Depth Anything V2 cues, produces dense metric depth maps on unseen ZJUL5, TUM, and Replica scenes.
    All supervision and Gaussian seeding use D_k from this network; no experiment verifies the network's zero-shot depth accuracy on these datasets independently of the SLAM loop.
  • domain assumption Relative camera poses from the tracking frontend are accurate enough for the multi-frame integration network to compute metric depth.
    DoD uses relative poses as metric-scale input, and these poses come from the same optimization that D_k supervises; the paper does not analyze failure of this feedback loop.
  • domain assumption The no-ToF DoD depth prediction is a trustworthy reference for rejecting sparse ToF outliers.
    Outlier handling (Sec. 3.2) discards any measurement whose deviation from the no-ToF prediction exceeds quantile q; if the prediction is biased, valid sensor data is discarded.
  • domain assumption Additive heteroscedastic Gaussian noise on Replica approximates real sparse-ToF noise well enough to support robustness conclusions.
    The noise robustness study (Fig. 6) uses a synthetic noise model, not real ZJUL5 noise statistics.
  • domain assumption Depth Anything V2 monocular depth provides scale-free shape cues that help DoD in textureless or low-parallax areas.
    Monocular cues are injected into the DoD encoder (Sec. 3.2); the method's gain depends on this pretrained prior.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ToF-Splatting: Dense SLAM using Sparse Time-of-Flight Depth and Multi-Frame Integration." pith.science (2026). https://pith.science/paper/AXIAAMZK

@misc{pith2026250416545,
  author       = {Pith},
  title        = {Pith review of: ToF-Splatting: Dense SLAM using Sparse Time-of-Flight Depth and Multi-Frame Integration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AXIAAMZK}},
  note         = {Machine review of arXiv:2504.16545}
}
read the original abstract

Time-of-Flight (ToF) sensors provide efficient active depth sensing at relatively low power budgets; among such designs, only very sparse measurements from low-resolution sensors are considered to meet the increasingly limited power constraints of mobile and AR/VR devices. However, such extreme sparsity levels limit the seamless usage of ToF depth in SLAM. In this work, we propose ToF-Splatting, the first 3D Gaussian Splatting-based SLAM pipeline tailored for using effectively very sparse ToF input data. Our approach improves upon the state of the art by introducing a multi-frame integration module, which produces dense depth maps by merging cues from extremely sparse ToF depth, monocular color, and multi-view geometry. Extensive experiments on both synthetic and real sparse ToF datasets demonstrate the viability of our approach, as it achieves state-of-the-art tracking and mapping performances on reference datasets.

Figures

Figures reproduced from arXiv: 2504.16545 by the authors.

Figure 1
Figure 1. Overview of our ToF-Splatting method. Our method combines sparse ToF depth, multi-view geometry from a buffer of keyframes, and monocular cues (left) to perform into a unique end-to-end dense SLAM framework enabled by a Gaussian Splatting. Abstract Time-of-Flight (ToF) sensors provide efficient active depth sensing at relatively low power budgets; among such designs, only very sparse measurements from low￾resolution… view at source ↗
Figure 2
Figure 2. ToF-Splatting Pipeline. Our method involves three main modules: a Tracking frontend estimating camera poses, a Multi-Frame Integration module that predicts dense depth maps from sparse ToF measurements and multi-view geometry, and a Mapping backend modeling the 3D scene representation via 3D Gaussian Splatting. represents the only attempt to exploit “lightweight”, sparse depth measurements within a dense SLAM system… view at source ↗
Figure 3
Figure 3. Qualitative results on the ZJUL5 dataset [19]. We show meshes obtained by fusing rendered depth maps with TSDF and marching cubes (left), and 3D trajectories (right) on 3 scenes selected from the ZJUL5 dataset [19]. and render depth and color from 3DGS. Then, we perform TSDF integration and extract the final mesh through march￾ing cubes. Following [19], the meshes are then evaluated by computing Accuracy, Completion… view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Replica Qualitatives. We provide qualitative results on Replica [28] to demonstrate the generalization capabilities of our method. On the left from top to bottom, meshes obtained respectively on scenes Office2 and Room2. On the right, details from the scene Room0. ToF-…
Figure 5
Figure 5. Figure 5: Impact of depth sparsity. We test on Replica [28] with different simulated depth sparsity levels to assess the capability to exploit higher input densities. MAE and ATE smoothly decrease, whereas rendering metrics appear to be less affected. 0.06 0.08 0.10 0.12 0.14 0%…
Figure 6
Figure 6. Figure 6: Impact of depth noise. We test on Replica [28] injecting different amounts of noise ξ ∈ [0.00, 0.01, 0.05, 0.10]. ToF￾Splatting demonstrates to be effective at dealing with noise, with error increasing almost linearly with the injected amount of noise. Splatting enable…
Figure 7
Figure 7. Figure 7: Qualitative results on TUM RGB-D dataset. Depth maps for fr1/office (left), fr2/xyz (center), fr3/desk (right) as esti￾mated by DoD [3]. a high density of outliers resulting from sparse Time-of￾Flight (ToF) data, as well as environmental difficulties such as poor textu…
Figure 8
Figure 8. Figure 8: Replica Qualitatives. We provide the trajectory and mesh reconstruction qualitative results on each scene provided by Replica. ToF-Splatting enables effective mesh reconstructions and accurate tracking. operate effectively even when the ToF sensor’s frame rate is signi…
Figure 9
Figure 9. Figure 9: ZJUL5 Qualitatives. We provide the trajectory and mesh reconstruction qualitative results on the scenes provided by ZJUL5. ToF-Splatting enables effective mesh reconstructions and accurate tracking. jection from one of the previous source views. This adjust￾ment aligns…
Figure 10
Figure 10. Figure 10: Impact of Temporal Sparsity. The three line plots represent respectively mean absolute error, F-score, and absolute trajectory error as the subsampling ratio of the ToF frames increases. As temporal sparsity grows, a gradual decline in overall performance is ob￾served…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

54 extracted references · 39 canonical work pages

  1. [1]

    Orb-slam3: An accu- rate open-source library for visual, visual–inertial, and mul- timap slam

    Carlos Campos, Richard Elvira, Juan J G ´omez Rodr´ıguez, Jos´e MM Montiel, and Juan D Tard´os. Orb-slam3: An accu- rate open-source library for visual, visual–inertial, and mul- timap slam. IEEE Transactions on Robotics , 37(6):1874– 1890, 2021. 2

    2021

  2. [2]

    Vcr-gaus: View consistent depth-normal regularizer for gaussian surface reconstruction

    Hanlin Chen, Fangyin Wei, Chen Li, Tianxin Huang, Yun- song Wang, and Gim Hee Lee. Vcr-gaus: View consistent depth-normal regularizer for gaussian surface reconstruction. arXiv preprint arXiv:2406.05774, 2024. 2, 5

    arXiv 2024

  3. [3]

    Depth on demand: Streaming dense depth from a low frame-rate active sensor

    Andrea Conti, Matteo Poggi, Valerio Cambareri, and Stefano Mattoccia. Depth on demand: Streaming dense depth from a low frame-rate active sensor. In European Conference on Computer Vision (ECCV), 2024. 2, 4, 1

    work page 2024

  4. [4]

    Mat- toccia

    Andrea Conti, Matteo Poggi, Valerio Cambareri, and S. Mat- toccia. Range-agnostic multi-view depth estimation with keyframe selection. 2024 International Conference on 3D Vision (3DV), pages 1350–1359, 2024. 4

    work page 2024

  5. [5]

    ScanNet: Richly-annotated 3d reconstructions of indoor scenes

    Angela Dai, Angel X Chang, Manolis Savva, Maciej Hal- ber, Thomas Funkhouser, and Matthias Nießner. ScanNet: Richly-annotated 3d reconstructions of indoor scenes. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 5828–5839, 2017. 4, 5, 2

    work page 2017

  6. [6]

    BundleFusion: Real-time globally consistent 3d reconstruction using on-the-fly surface reintegration

    Angela Dai, Matthias Nießner, Michael Zollh ¨ofer, Shahram Izadi, and Christian Theobalt. BundleFusion: Real-time globally consistent 3d reconstruction using on-the-fly surface reintegration. ACM Transactions on Graphics (ToG), 36(4): 1, 2017. 1, 2, 6, 7

    work page 2017

  7. [7]

    Plgslam: Progressive neural scene represenation with local to global bundle adjustment

    Tianchen Deng, Guole Shen, Tong Qin, Jianyu Wang, Wen- tao Zhao, Jingchuan Wang, Danwei Wang, and Weidong Chen. Plgslam: Progressive neural scene represenation with local to global bundle adjustment. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 19657–19666, 2024. 2

    work page 2024

  8. [8]

    Keyframe-based real-time camera tracking

    Zilong Dong, Guofeng Zhang, Jiaya Jia, and Hujun Bao. Keyframe-based real-time camera tracking. In Proceedings of the IEEE/CVF International Conference on Computer Vi- sion, pages 1538–1545. IEEE, 2009. 1

    work page 2009

Show all 54 references
  1. [9]

    Sugar: Surface- aligned gaussian splatting for efficient 3d mesh reconstruc- tion and high-quality mesh rendering

    Antoine Gu ´edon and Vincent Lepetit. Sugar: Surface- aligned gaussian splatting for efficient 3d mesh reconstruc- tion and high-quality mesh rendering. CVPR, 2024. 2

    2024

  2. [10]

    KinectFusion: real-time 3d reconstruction and interac- tion using a moving depth camera

    Shahram Izadi, David Kim, Otmar Hilliges, David Molyneaux, Richard Newcombe, Pushmeet Kohli, Jamie Shotton, Steve Hodges, Dustin Freeman, Andrew Davison, et al. KinectFusion: real-time 3d reconstruction and interac- tion using a moving depth camera. In Proceedings of the 24th ...

    2011

  3. [11]

    Eslam: Efficient dense slam system based on hybrid representation of signed distance fields

    Mohammad Mahdi Johari, Camilla Carta, and Franc ¸ois Fleuret. Eslam: Efficient dense slam system based on hybrid representation of signed distance fields. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 17408–17419, 2023. 2

    2023

  4. [12]

    Splatam: Splat track & map 3d gaussians for dense rgb-d slam

    Nikhil Keetha, Jay Karhade, Krishna Murthy Jatavallabhula, Gengshan Yang, Sebastian Scherer, Deva Ramanan, and Jonathon Luiten. Splatam: Splat track & map 3d gaussians for dense rgb-d slam. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition,...

    2024

  5. [13]

    3d gaussian splatting for real-time radiance field rendering

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimkuehler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering. ACM Transactions on Graphics (TOG), 2023. 1, 2, 3

    2023

  6. [14]

    Parallel tracking and map- ping for small AR workspaces

    Georg Klein and David Murray. Parallel tracking and map- ping for small AR workspaces. In Proceedings of the IEEE and ACM International Symposium on Mixed and Aug- mented Reality, pages 225–234. IEEE, 2007. 1

    2007

  7. [15]

    Structure-slam: Low-drift monocular slam in indoor environments

    Yanyan Li, Nikolas Brasch, Yida Wang, Nassir Navab, and Federico Tombari. Structure-slam: Low-drift monocular slam in indoor environments. IEEE Robotics and Automa- tion Letters, 5(4):6583–6590, 2020. 2

    2020

  8. [16]

    Deltar: Depth estimation from a light-weight tof sensor and rgb image

    Yijin Li, Xinyang Liu, Wenqian Dong, Han Zhou, Hujun Bao, Guofeng Zhang, Yinda Zhang, and Zhaopeng Cui. Deltar: Depth estimation from a light-weight tof sensor and rgb image. In European Conference on Computer Vision ,

  9. [17]

    Loopy-slam: Dense neural slam with loop closures

    Lorenzo Liso, Erik Sandstr ¨om, Vladimir Yugay, Luc Van Gool, and Martin R Oswald. Loopy-slam: Dense neural slam with loop closures. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 20363–20373, 2024. 2

    2024

  10. [18]

    Robust keyframe-based monocular SLAM for augmented reality

    Haomin Liu, Guofeng Zhang, and Hujun Bao. Robust keyframe-based monocular SLAM for augmented reality. In Proceedings of the IEEE International Symposium on Mixed and Augmented Reality, pages 1–10. IEEE, 2016. 1

    2016

  11. [19]

    Multi-modal neural radiance field for monocular dense slam with a light-weight tof sensor

    Xinyang Liu, Yijin Li, Yanbin Teng, Hujun Bao, Guofeng Zhang, Yinda Zhang, and Zhaopeng Cui. Multi-modal neural radiance field for monocular dense slam with a light-weight tof sensor. 2023 IEEE/CVF International Conference on Computer Vision (ICCV), pages 1–11, 2023. 2, 4, 5, ...

    2023

  12. [20]

    Gregor Luetzenburg, Aart Kroon, and Anders A. Bjørk. Evaluation of the Apple iPhone 12 Pro LiDAR for an Ap- plication in Geosciences. Scientific Reports, 11(1), 2021. 1

    2021

  13. [21]

    Dynamic 3d gaussians: Tracking by persis- tent dynamic view synthesis

    Jonathon Luiten, Georgios Kopanas, Bastian Leibe, and Deva Ramanan. Dynamic 3d gaussians: Tracking by persis- tent dynamic view synthesis. 2024 International Conference on 3D Vision (3DV), pages 800–809, 2023. 2

    2024

  14. [22]

    Kelly, and An- drew J

    Hidenobu Matsuki, Riku Murai, Paul H.J. Kelly, and An- drew J. Davison. Gaussian splatting slam. 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 18039–18048, 2023. 2, 4, 5, 6, 7, 8

    2024

  15. [23]

    NeRF: Representing scenes as neural radiance fields for view syn- thesis

    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. In Proceedings of the European Conference on Com- puter Vision, pages 405–421. Springer, 2020. 1

    2020

  16. [24]

    Tard ´os

    Ra ´ul Mur-Artal and Juan D. Tard ´os. ORB-SLAM2: An Open-Source SLAM System for Monocular, Stereo, and RGB-D Cameras. IEEE Transactions on Robotics , 33(5): 1255–1262, 2017. 2

    2017

  17. [25]

    Rgb guided tof imag- 9 ing system: A survey of deep learning-based methods

    Xin Qiao, Matteo Poggi, Pengchao Deng, Hao Wei, Chenyang Ge, and Stefano Mattoccia. Rgb guided tof imag- 9 ing system: A survey of deep learning-based methods. In- ternational Journal of Computer Vision , pages 1–38, 2024. 2

    2024

  18. [26]

    Point-slam: Dense neural point cloud-based slam

    Erik Sandstr ¨om, Yue Li, Luc Van Gool, and Martin R Os- wald. Point-slam: Dense neural point cloud-based slam. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 18433–18444, 2023. 2

    2023

  19. [27]

    Bad slam: Bundle adjusted direct rgb-d slam

    Thomas Schops, Torsten Sattler, and Marc Pollefeys. Bad slam: Bundle adjusted direct rgb-d slam. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 134–144, 2019. 2

    2019

  20. [28]

    Julian Straub, Thomas Whelan, Lingni Ma, Yufan Chen, Erik Wijmans, Simon Green, Jakob J. Engel, Raul Mur-Artal, Carl Ren, Shobhit Verma, Anton Clarkson, Mingfei Yan, Brian Budge, Yajie Yan, Xiaqing Pan, June Yon, Yuyang Zou, Kimberly Leon, Nigel Carter, Jesus Briales, Tyler Gi...

    1906 arXiv

  21. [29]

    A benchmark for the eval- uation of RGB-D SLAM systems

    J ¨urgen Sturm, Nikolas Engelhard, Felix Endres, Wolfram Burgard, and Daniel Cremers. A benchmark for the eval- uation of RGB-D SLAM systems. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, pages 573–580. IEEE, 2012. 5, 6

    2012

  22. [30]

    iMAP: Implicit mapping and positioning in real-time

    Edgar Sucar, Shikun Liu, Joseph Ortiz, and Andrew J Davi- son. iMAP: Implicit mapping and positioning in real-time. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 6229–6238, 2021. 2, 6, 7, 8

    2021

  23. [31]

    Cnn-slam: Real-time dense monocular slam with learned depth prediction

    Keisuke Tateno, Federico Tombari, Iro Laina, and Nassir Navab. Cnn-slam: Real-time dense monocular slam with learned depth prediction. In Proceedings of the IEEE con- ference on computer vision and pattern recognition , pages 6243–6252, 2017. 2

    2017

  24. [32]

    Droid-SLAM: Deep vi- sual SLAM for monocular, stereo, and RGB-D cameras

    Zachary Teed and Jia Deng. Droid-SLAM: Deep vi- sual SLAM for monocular, stereo, and RGB-D cameras. Advances in Neural Information Processing Systems , 34: 16558–16569, 2021. 2, 7

    2021

  25. [33]

    Oswald, and Matteo Poggi

    Fabio Tosi, Youmin Zhang, Ziren Gong, Erik Sandstr ¨om, Stefano Mattoccia, Martin R. Oswald, and Matteo Poggi. How nerfs and 3d gaussian splatting are reshaping slam: a survey, 2024. 1, 2

    2024

  26. [34]

    Co- slam: Joint coordinate and sparse parametric encodings for neural real-time slam

    Hengyi Wang, Jingwen Wang, and Lourdes Agapito. Co- slam: Joint coordinate and sparse parametric encodings for neural real-time slam. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 13293–13302, 2023. 2

    2023

  27. [35]

    Differentiable surface splatting for point-based geometry processing

    Yifan Wang, Felice Serena, Shihao Wu, Cengiz ¨Oztireli, and Olga Sorkine-Hornung. Differentiable surface splatting for point-based geometry processing. ACM Transactions on Graphics (TOG), 38:1 – 14, 2019. 3

    2019

  28. [36]

    ElasticFusion: Real-time dense SLAM and light source estimation

    Thomas Whelan, Renato F Salas-Moreno, Ben Glocker, An- drew J Davison, and Stefan Leutenegger. ElasticFusion: Real-time dense SLAM and light source estimation. The International Journal of Robotics Research , 35(14):1697– 1716, 2016. 1, 2, 6, 7

    2016

  29. [37]

    Kn-slam: Keypoints and neural implicit en- coding slam

    Xingming Wu, Zimeng Liu, Yuxin Tian, Zhong Liu, and Weihai Chen. Kn-slam: Keypoints and neural implicit en- coding slam. IEEE Transactions on Instrumentation and Measurement, 73:1–12, 2024. 2

    2024

  30. [38]

    Multi-modal neural radiance field for monocular dense slam with a light- weight tof sensor

    Liu Xinyang, Li Yijin, Teng Yanbin, Bao Hujun, Zhang Guofeng, Zhang Yinda, and Cui Zhaopeng. Multi-modal neural radiance field for monocular dense slam with a light- weight tof sensor. In International Conference on Computer Vision (ICCV), 2023. 2

    2023

  31. [39]

    Gs-slam: Dense visual slam with 3d gaussian splatting

    Chi Yan, Delin Qu, Dan Xu, Bin Zhao, Zhigang Wang, Dong Wang, and Xuelong Li. Gs-slam: Dense visual slam with 3d gaussian splatting. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 19595–19604, 2024. 2

    2024

  32. [40]

    Depth any- thing v2

    Lihe Yang, Bingyi Kang, Zilong Huang, Zhen Zhao, Xiao- gang Xu, Jiashi Feng, and Hengshuang Zhao. Depth any- thing v2. ArXiv, abs/2406.09414, 2024. 4, 8

    2024 arXiv

  33. [41]

    Os- wald

    Vladimir Yugay, Yue Li, Theo Gevers, and Martin R. Os- wald. Gaussian-slam: Photo-realistic dense slam with gaus- sian splatting. ArXiv, abs/2312.10070, 2023. 5

    2023 arXiv

  34. [42]

    Os- wald

    Vladimir Yugay, Yue Li, Theo Gevers, and Martin R. Os- wald. Gaussian-slam: Photo-realistic dense slam with gaus- sian splatting, 2023. 2

    2023

  35. [43]

    Recovering consistent video depth maps via bundle opti- mization

    Guofeng Zhang, Jiaya Jia, Tien-Tsin Wong, and Hujun Bao. Recovering consistent video depth maps via bundle opti- mization. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1–8. IEEE,

  36. [44]

    Go-slam: Global optimization for consistent 3d in- stant reconstruction

    Youmin Zhang, Fabio Tosi, Stefano Mattoccia, and Matteo Poggi. Go-slam: Global optimization for consistent 3d in- stant reconstruction. In Proceedings of the IEEE/CVF Inter- national Conference on Computer Vision, pages 3727–3737,

  37. [45]

    Nice-slam: Neural implicit scalable encoding for slam

    Zihan Zhu, Songyou Peng, Viktor Larsson, Weiwei Xu, Hu- jun Bao, Zhaopeng Cui, Martin R Oswald, and Marc Polle- feys. Nice-slam: Neural implicit scalable encoding for slam. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 12786–1279...

    2022

  38. [46]

    NICE-SLAM: Neural implicit scalable encoding for SLAM

    Zihan Zhu, Songyou Peng, Viktor Larsson, Weiwei Xu, Hu- jun Bao, Zhaopeng Cui, Martin R Oswald, and Marc Polle- feys. NICE-SLAM: Neural implicit scalable encoding for SLAM. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 12786– 1279...

    2022

  39. [47]

    Oswald, Andreas Geiger, and Marc Pollefeys

    Zihan Zhu, Songyou Peng, Viktor Larsson, Zhaopeng Cui, Martin R. Oswald, Andreas Geiger, and Marc Pollefeys. Nicer-slam: Neural implicit scene encoding for rgb slam. 2024 International Conference on 3D Vision (3DV) , pages 42–52, 2023. 4

    2024

  40. [48]

    Ogni-dc: Robust depth com- pletion with optimization-guided neural iterations

    Yiming Zuo and Jia Deng. Ogni-dc: Robust depth com- pletion with optimization-guided neural iterations. arXiv preprint arXiv:2406.11711, 2024. 6, 8

    2024 arXiv

  41. [49]

    ToF-Splatting: Dense SLAM using Sparse Time-of-Flight Depth and Multi-Frame Integration

    Matthias Zwicker, Hanspeter Pfister, Jeroen van Baar, and Markus H. Gross. Surface splatting. Proceedings of the 28th annual conference on Computer graphics and interac- tive techniques, 2001. 3 10 ToF-Splatting: Dense SLAM using Sparse Time-of-Flight Depth and Multi-Frame Int...

    2001

  42. [50]

    Indeed, we obtain high-quality depth maps from sparse inputs

    TUM RGB-D DoD Qualitative Results In Figure 7 we present three pairs of color views and re- constructed depth maps to assess that our method [3], as integrated in the ToF-splatting pipeline, generalizes well to unseen datasets such as TUM RGB-D at test time. Indeed, we obtain ...

  43. [51]

    To achieve this, we first fit the entire scene and render depth and color images for each pose estimated by ToF-Splatting

    Replica Qualitative Results In Figure 8, we present the reconstructed mesh and the pre- dicted trajectory for each scene in the Replica [28] dataset. To achieve this, we first fit the entire scene and render depth and color images for each pose estimated by ToF-Splatting. Thes...

  44. [52]

    ZJUL5 Qualitative Results Figure 9 illustrates the reconstructed meshes and predicted trajectories for the scenes included in the ZJUL5 [19] dataset. Unlike the Replica dataset [28], which primar- ily focuses on synthetic environments, the ZJUL5 dataset presents real-world sce...

  45. [53]

    Thus, only a subset of the RGB frames is coupled with sparse depth in- formation

    Temporal Sparsity Finally, we study ToF-Splatting performance under tempo- ral sparsity, which refers to scenarios where the ToF sensor frame rate is lower than that of the RGB camera. Thus, only a subset of the RGB frames is coupled with sparse depth in- formation. Such a sit...

  46. [54]

    Specifically, we significantly modify its in- nermost logic to integrate monocular cues and handle a larger number of frames to overcome the original two-frame configuration

    Depth on Demand Training Details In ToF-Splatting, we perform multi-frame integration adapting the Depth on Demand framework [3] to our spe- cific use case. Specifically, we significantly modify its in- nermost logic to integrate monocular cues and handle a larger number of fr...

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.