Pith. sign in

REVIEW 4 major objections 10 minor 1 cited by

Uni-SLAM: Uncertainty-Aware Neural Implicit SLAM for Real-Time Dense Indoor Scene Reconstruction

T0 review · 4 major / 10 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 Uni-SLAM shows that weighting each pixel by a volume-rendering confidence score improves neural implicit SLAM tracking and mapping without giving up real-time operation.

desk verdict Solid, well-engineered SLAM systems paper with a genuinely new uncertainty heuristic; the main gap is that the heuristic is never stress-tested under pose failure. read the letter →

arxiv 2412.00242 v1 pith:YLD3FHIW submitted 2024-11-29 cs.CV

Shaoxiang Wang , Yaxu Xie , Chun-Peng Chang , Christen Millerdurai , Alain Pagani , Didier Stricker This is my paper · ORCID
classification cs.CV
keywords neuralimplicitSLAMdenseRGB-Dreconstructionpredictiveuncertaintyvolumerenderingterminationprobabilityhashgridscenerepresentationbundleadjustmentindoorreal-timemapping
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

Uni-SLAM sets out to show that a dense neural implicit SLAM system should not treat all RGB-D frames and pixels as equally trustworthy. Its central proposal is a cheap, model-free predictive uncertainty computed from the volume-rendering termination probability, which reweights the tracking and mapping losses so that poorly observed pixels count for less. The same image-level uncertainty triggers additional local bundle adjustment when the camera enters unfamiliar territory, instead of relying only on fixed-frequency global mapping. On the Replica, ScanNet, and TUM RGB-D benchmarks the paper reports state-of-the-art tracking accuracy and reconstruction quality among the compared neural implicit systems, with a 25% lower depth L1 error and a 66.86% sub-centimeter completion ratio on Replica, while running at about 8 FPS. If correct, this establishes uncertainty-aware selective optimization, rather than uniform fixed-frequency mapping, as the better design for real-time implicit indoor SLAM.

What carries the argument

Central machinery is the termination-probability confidence signal. Along a ray with sample weights $w_i = T_i(1-\exp(-\sigma(x_i)))$, the accumulated value $p(r)=\sum_{i=1}^N w_i$ telescopes to $1-\exp(-\sum_i\sigma(x_i))$, bounding it in $(0,1)$ by construction. The paper treats $p(r)$ as the probability that the ray terminates on a known surface, defines $\beta_m=(1-p(r_m))^2$ for each pixel and $\beta=\frac{1}{M}\sum_m\beta_m$ per image, and converts $\beta_m$ into a binary mask $CF_m=\mathbf 1[\beta_m\le \beta^\text{unc}_m]$ that multiplies the SDF, depth, and tracking color losses. This is what lets the system discard outliers and emphasize reliable data during pose optimization. The second load-bearing mechanism is decoupled geometry and appearance hash grids, which allow high-frequency color and geometry to be sampled at different resolutions, plus the uncertainty-gated local-to-global bundle adjustment schedule that adds local BA when $\beta>\beta^\text{unc}$ and LLCO when co-visibility exceeds 0.95.

What would settle it

Use a Replica or ScanNet sequence with known invalid-depth regions, run Uni-SLAM, and record per-pixel $p(r)$, $\beta_m$, and depth error against the sensor depth. The claim predicts that pixels with $\beta_m$ below threshold have small depth error and that high-$\beta_m$ pixels coincide with invalid or unobserved depth; if any low-uncertainty pixel shows depth error larger than the truncation distance, or if the uncertainty map does not flag deliberately corrupted depth regions, the confidence proxy is falsified.

Watch

Extended reading notes

Core claim

The paper's claim is that the accumulated termination probability of a sampled ray, $p(r)=\sum_i w_i = 1-\exp(-\sum_i \sigma(x_i))$, is a usable online measure of how well the current model knows the observed surface: near 1 for well-observed, accurately tracked pixels and near 0 for invalid-depth, occluded, or never-seen regions. From it Uni-SLAM derives pixel-level uncertainty $\beta_m=(1-p(r_m))^2$ and image-level uncertainty $\beta=\frac{1}{M}\sum_m \beta_m$, uses a thresholded confidence mask $CF_m$ to reweight SDF, depth, and tracking color losses, and leaves the mapping color loss unweighted so RGB can compensate for invalid depth. A decoupled pair of multiresolution hash grids represents geometry and appearance separately, and image-level uncertainty plus a co-visibility check activates local bundle adjustment and loop-closure optimization on top of constant global BA. The reported outcome is improved pose estimation in every Replica scene, lower ATE on ScanNet and TUM RGB-D than the compared neural implicit SLAM systems, thinner structures in the reconstructed meshes, and real-time operation at roughly 8.4 FPS.

Load-bearing premise

The system trusts that the accumulated termination probability is near 1 precisely for well-observed, accurately posed pixels and near 0 for invalid or unseen regions, so a confidently wrong model could be reweighted up instead of filtered.

Editorial extensions

If this is right

  • If the results hold, a neural implicit SLAM system can gain accuracy by computing a per-pixel confidence from the model's own rendering weights, with no extra uncertainty network and no additional training cost.
  • Tracking on Replica improves to a 0.45 cm average ATE RMSE, with ScanNet and TUM RGB-D averages of 7.01 cm and 2.05 cm, both best among the neural implicit baselines compared.
  • Mapping quality on Replica improves to 0.89 cm depth L1, 0.92 cm accuracy and completion, and 66.86% completion within 1 cm, corresponding to thinner structures such as chair legs and table edges.
  • The uncertainty-driven mapping trigger is not tied to Uni-SLAM's own representation: applying it to BSLAM improves that system's ATE from 1.19 cm to 1.07 cm and its completion ratio from 57.18% to 58.36%.
  • Real-time operation is preserved: roughly 8.4 FPS on Replica with an RTX 4090, comparable to Co-SLAM and faster than ESLAM, Nice-SLAM, and BSLAM in the reported setup.

Reading between the lines

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

  • Editorial extension: because $\beta$ is computed from the current model state, it could double as an online alarm for motion blur, sensor dropout, or aggressive camera motion, enabling the SLAM system to slow down, request new views, or refuse bad poses before drift accumulates.
  • Editorial extension: the same uncertainty could drive active next-best-view selection, since high-$\beta$ rays mark regions the model has not yet observed; the paper mentions active sampling as future work but does not test it.
  • Editorial extension: the hand-set thresholds $\beta^\text{unc}_m$ and $\beta^\text{unc}$ vary across datasets, so a principled automatic calibration, such as normalizing by the per-image uncertainty distribution, would be needed for deployment outside the tuned benchmarks.
  • Editorial extension: the decision not to reweight color during mapping rests on RGB compensating for invalid depth; in a sensor with reliable depth, reweighting color in mapping might become beneficial, which could be tested directly.
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 / 10 minor

Summary. The manuscript proposes Uni-SLAM, a real-time dense neural implicit RGB-D SLAM system with three main components: (i) a decoupled representation with separate multiresolution hash grids for geometry (SDF) and appearance, both decoded by tiny MLPs and rendered by standard volume rendering; (ii) a model-free "predictive uncertainty" defined per pixel as beta_m = (1 - p(r))^2, where p(r) is the accumulated termination probability along the depth-guided sampling ray, used to hard-mask and reweight the SDF, depth, and (during tracking only) color loss terms; and (iii) an uncertainty-triggered mapping schedule combining fixed-frequency global bundle adjustment, image-level-uncertainty-activated local BA, and a co-visibility-based local loop closure optimization (LLCO). The system is evaluated on Replica, ScanNet, and TUM RGB-D against Nice-SLAM, Co-SLAM, ESLAM, BSLAM, and others, reporting average ATE RMSE of 0.45 cm on Replica (vs. 0.63 cm for ESLAM), Depth L1 of 0.89 cm, 66.86% completion within 1 cm, and PSNR of 31.62 dB, with a claimed 25% reduction in depth L1 error over the prior state of the art, together with a runtime of about 8.4 FPS on Replica. The supplement includes per-scene tables, ablations (representation choice, reweighting terms, BA components, hash grid sizes, Gaussian-assumption uncertainty), a telescoping-sum proof for the termination probability, and a plug-in experiment showing the strategic BA transfers to BSLAM.

Significance. If the reported results hold up, Uni-SLAM is a meaningful advance: it is the first dense implicit RGB-D SLAM system to use the rendering termination probability as a free, model-free confidence signal, and its benchmark numbers (best average ATE on all three datasets among the compared implicit methods) support the headline claims. Strengths worth crediting explicitly: baseline results are reproduced from open-source code with five runs (median reported) under a consistent mesh-culling protocol; the termination-probability identity is derived correctly as a telescoping sum; ablations substantiate the decoupled representation, the choice of reweighting terms, and the value of LLCO; and the BSLAM plug-in experiment demonstrates transfer of the BA strategy.

major comments (4)
  1. [Sec. 3.2, Eqs. (4)-(7); Fig. 3; Sec. 4.3] The central mechanism claim -- that beta_m = (1 - p(r))^2 identifies unreliable observations and that hard masking via Eq. (7) "progressively filters out outliers" -- is validated only in the regime where the model is uncertain because of missing coverage (invalid depth in Fig. 3, unseen areas in Supp. Fig. 10). The complementary regime is untested: after a pose perturbation, depth-guided samples can land on a previously learned surface, giving p(r) near 1 and CF_m = 1, so the inconsistent observation is up-weighted instead of filtered, which can amplify drift in the joint tracking/mapping loop. The supplement proof (Sec. A.2) establishes only the identity p(r) = 1 - exp(-sum sigma); it does not establish the semantic claim that p(r) is large exactly when the observation is reliable under pose error. I request a controlled experiment: inject pose perturbations of increasing magnitude during tracking (with the map frozen), report the resulting distribution of beta_m and the post-optimization tracking error, and compare against a baseline that masks only pixels with explicitly invalid depth. This would show whether the reweighting provides information beyond a standard depth-validity mask and whether the aggregate gains in Fig. 9 arise from the proposed mechanism.
  2. [Supp. Tab. 14; Tab. 2; Fig. 6] Several main-text numbers do not reconcile with the per-scene supplement tables. (i) The per-scene Depth L1 values for Ours in Tab. 14 (0.81, 0.77, 1.13, 0.70, 1.11, 1.52, 1.15, 0.99) average to 1.02 cm, yet both Tab. 14 and Tab. 2 report 0.89 cm; the headline claim of a 25% reduction in Depth L1 over ESLAM (1.18 cm) depends on this figure. (ii) The ESLAM completion ratio at 1 cm is 62.25% as the mean of the Tab. 14 per-scene values but is reported as 63.99% in Tab. 2. (iii) For Ours on office-3, Fig. 6 reports a 50.19% completion ratio at 1 cm while Tab. 14 lists 50.91%. Please reconcile these numbers or state clearly which runs or configurations each table refers to.
  3. [Sec. 4.1, Tabs. 1-4] The quantitative comparison is statistically asymmetric: for the baselines the paper reports the median of five runs, while the proposed method is reported as a single run without error bars. Some of the claimed margins are modest (TUM average ATE: 2.05 vs. 2.14 for ESLAM; completion ratio at 1 cm on Replica: 66.86 vs. 63.99), so the claim that Uni-SLAM "significantly improves" over current methods is not yet substantiated. Please report mean +/- std (or per-run values) for the proposed method over multiple runs for the main tables, and mark which differences are robust to run-to-run variation.
  4. [Supp. Sec. A.1; Sec. 3.4] The uncertainty pipeline relies on several hand-set, dataset-specific thresholds: beta_unc_m = 1e-2 for pixel masking, beta_unc = 1e-3 (2e-3 on TUM) for activating local BA, mapping frequency n = 4 (Replica, TUM) or 5 (ScanNet), and tau_cov = 0.95 for loop closure. Because Eq. (7) is a hard 0/1 mask, behavior is discontinuous in beta_unc_m, and no sensitivity analysis is provided. I ask for a sweep over these thresholds (at least beta_unc_m and beta_unc) on one Replica and one TUM sequence, and for a statement of which thresholds were used in the BSLAM plug-in experiment (Tab. 10), to establish that the reported gains do not depend critically on tuning.
minor comments (10)
  1. [Tab. 1] In the MIPS-Fusion row, the Off-2 entry reads "2020", which appears to be a typo (likely 2.02 or 20.20); please correct it.
  2. [Tabs. 5 and 12] The tables label the BSLAM baseline as "BSLAM [47]" in several places (e.g., Tab. 5 and Tab. 12); BSLAM is reference [20], while [47] is Point-SLAM.
  3. [Supp. Tab. 12] In the Replica block, the Nice-SLAM mapping column reads "29.3 x 0", which appears to be a typo for "29.3 x 60" (cf. Tab. 5).
  4. [Eq. (7)] The notation CF_m = 1(1 - beta_m) is confusing; write the indicator function explicitly, e.g., CF_m = 1{beta_m <= beta_unc_m}.
  5. [Sec. 3.2] "never unobserved regions" should read "never observed regions".
  6. [Supp. Figs. 17 and 19] The figure legends contain a leftover template string ("SA T [8] BUTD-DETR [8] ViewRefer [8]") that should be removed.
  7. [Algorithm 1] The comment "Optimazation" should be corrected to "Optimization".
  8. [Abstract and Conclusion] The claims of "real-time performance" should be qualified: per Supp. Tab. 12 the method runs at 8.37 FPS on Replica, 4.88 FPS on ScanNet, and only 2.7 FPS on TUM RGB-D, which is below the real-time threshold.
  9. [Abstract and Related Work] The abstract's statement that previous methods "do not consider varying quality of input RGB-D data" is too strong given that UNCLE-SLAM [48] is cited in the related work; please soften or qualify it.
  10. [Fig. 9] The reweighting ablation reports results on "TUM RGB-D" without identifying the specific sequence; please specify which sequence(s) were used.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the uncertainty-guided reweighting is an empirically tested heuristic, and the headline SLAM claims are measured against external benchmarks.

full rationale

Uni-SLAM's central claims are tracking and reconstruction accuracy on Replica, ScanNet, and TUM RGB-D, measured against independent baselines with standard metrics (ATE RMSE, Depth L1, accuracy, completion ratio, PSNR). These claims do not reduce to the paper's own definitions or fitted values. The pixel-level uncertainty beta_m = (1-p(r))^2 is not fitted to the target metrics; it is a deterministic function of the volume-rendering termination probability p(r), an existing quantity in NeRF-style rendering. The paper explicitly labels the link between beta_m and data quality as a hypothesis: 'We hypothesize that pixels with invalid depth or motion blur, caused by sensor issues or sudden motion changes, should exhibit high uncertainty, while well-observed regions should display low uncertainty.' It then tests this hypothesis through ablations (Fig. 9, Tab. 6, Tab. 7) and a plug-in experiment on BSLAM (Tab. 10), which provides an independent check that the strategy transfers. Using a model-derived confidence to reweight that same model's loss is self-referential as a training mechanism, but it is not definitional circularity: the final benchmark numbers come from ground-truth poses and meshes, not from beta_m, and the reweighting could have degraded rather than improved performance. The per-dataset thresholds (beta_unc_m=1e-2, beta_unc=1e-3 or 2e-3 on TUM) are hand-set hyperparameters, not parameters fitted to a subset of data and then renamed as predictions. No load-bearing self-citations or imported uniqueness theorems appear; cited works such as NeRF, DENS, and BayesRays are external prior art. The proof in Supp. A.2 only establishes the telescoping identity sum w_n = 1 - exp(-sum sigma), which is a mathematical property of the rendering weights, not a validation that the uncertainty measure tracks ground-truth data quality. The derivation chain is therefore self-contained; the main vulnerability is an empirical assumption about what p(r) indicates, which is a correctness or robustness concern rather than circularity.

Assumptions & free parameters 7 free parameters · 5 assumptions · 1 invented entities

The main free parameters are the uncertainty thresholds and mapping frequency, which are tuned per dataset and directly control loss masking and BA activation. The key axioms are the validity of termination probability as a data-quality measure, the depth-guided sampling assumption, and the use of a ground-truth first pose. No new physical entities are introduced; the 'predictive uncertainty' is a model-derived scalar.

free parameters (7)
  • Pixel-level uncertainty threshold β_unc_m = 1e-2
    Used in CF_m (Eq. 7) to mask uncertain pixels during loss computation; fixed to 1e-2 without sensitivity analysis.
  • Image-level uncertainty threshold β_unc = 1e-3 (Replica), 2e-3 (TUM)
    Triggers additional local BA; tuned per dataset as documented in Supp A.1.
  • Mapping frequency n = 4 frames (Replica/TUM), 5 frames (ScanNet)
    Constant global BA interval; selected per dataset to balance speed and accuracy.
  • Co-visibility threshold τ_cov = 0.95
    Decides loop closure detection in Eq. 16; a minimum of 100 frames is also enforced.
  • Truncation distance τ_tr = 6 cm
    Defines the SDF truncation band in Sec. 3.3; inherited from prior SDF-based SLAM.
  • Loss weights λ_rgb, λ_dep, λ_sdf = Various (e.g., λ_rgb=5, λ_dep=0.1..1, λ_sdf=5..200)
    Hand-set separately for tracking and mapping; listed in Supp A.1.
  • Hash grid levels L = 16 geometry, 16 or 19 appearance
    Ablated in Supp B.1; Replica uses L=19 for appearance.
assumptions (5)
  • standard math Telescoping sum identity: sum_i w_i = 1 - exp(-sum_i sigma_i)
    Proof in Supp A.2 is a correct telescoping sum within the discrete volume rendering model.
  • domain assumption Well-observed regions have cumulative density such that p(r) is near 1; unobserved or invalid-depth regions have p(r) near 0
    Core assumption for the uncertainty definition in Sec. 3.2; not guaranteed when the model hallucinates geometry or pose is poor.
  • domain assumption Depth-guided sampling with approximate depth for invalid pixels still yields useful rays for uncertainty estimation
    Used in Fig. 3 to compute uncertainty for invalid-depth pixels via an approximate depth value.
  • domain assumption First camera pose is available (ground truth)
    Algorithm 1 in Supp initializes P1 with ground truth; the main paper does not analyze the impact of this choice.
  • domain assumption Co-visibility overlap computed from current pose estimates reliably detects loop closures
    Eq. 16 and Supp A.3; false positives could trigger unnecessary or harmful LLCO.
invented entities (1)
  • Predictive uncertainty β (pixel-level and image-level)
    purpose: Reweight loss functions and trigger additional bundle adjustment based on model confidence
    Defined purely from the model's own volume rendering weights; no external calibration or ground-truth uncertainty map is used.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Uni-SLAM: Uncertainty-Aware Neural Implicit SLAM for Real-Time Dense Indoor Scene Reconstruction." pith.science (2026). https://pith.science/paper/YLD3FHIW

@misc{pith2026241200242,
  author       = {Pith},
  title        = {Pith review of: Uni-SLAM: Uncertainty-Aware Neural Implicit SLAM for Real-Time Dense Indoor Scene Reconstruction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YLD3FHIW}},
  note         = {Machine review of arXiv:2412.00242}
}
read the original abstract

Neural implicit fields have recently emerged as a powerful representation method for multi-view surface reconstruction due to their simplicity and state-of-the-art performance. However, reconstructing thin structures of indoor scenes while ensuring real-time performance remains a challenge for dense visual SLAM systems. Previous methods do not consider varying quality of input RGB-D data and employ fixed-frequency mapping process to reconstruct the scene, which could result in the loss of valuable information in some frames. In this paper, we propose Uni-SLAM, a decoupled 3D spatial representation based on hash grids for indoor reconstruction. We introduce a novel defined predictive uncertainty to reweight the loss function, along with strategic local-to-global bundle adjustment. Experiments on synthetic and real-world datasets demonstrate that our system achieves state-of-the-art tracking and mapping accuracy while maintaining real-time performance. It significantly improves over current methods with a 25% reduction in depth L1 error and a 66.86% completion rate within 1 cm on the Replica dataset, reflecting a more accurate reconstruction of thin structures. Project page: https://shaoxiang777.github.io/project/uni-slam/

Figures

Figures reproduced from arXiv: 2412.00242 by the authors.

Figure 1
Figure 1. The reconstructed 3D mesh on the TUM RGB-D [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Uni-SLAM Architecture Overview. Our framework consists of two threads, tracking and mapping. While tracking is performed every frame for RGB-D stream, besides constant mapping is performed every n frame constantly with global BA, activated additional mapping process is executed to capture local scene information based on uncertainty and co-visibility check with local BA and local loop closure optimization (LLCO). Ou… view at source ↗
Figure 3
Figure 3. Termination Probability and Uncertainty. This figure illustrates the termination probability and uncertainty during ray sampling. For pixel A with valid depth (sampling by Ray 1), the sampling density is high along this ray, leading to a high termination probability and lower uncertainty. In contrast, for pixel B with invalid depth (sampling by Ray 2), the sampling density is low along this ray, resulting in a lower… view at source ↗
Figures from the paper (22 more)
Figure 5
Figure 5. Figure 5: Activated additional local BA. From position Pi to Pi+1, sudden large movements lead to difficulties in pose estimation and increased uncertainty due to unseen areas. The initialization of Init Pi+1 based on the constant speed assumption is hard to optimize. Therefore,…
Figure 6
Figure 6. Figure 6: Mesh Evaluation on Replica [53]. Our method outstands with its thin geometry details and higher texture fidelity compared to Co-SLAM [62] and BSLAM [20]. For example, the table and vase in room-0; the thinner office desk, chair backrest, and detailed reconstructed chai…
Figure 7
Figure 7. Figure 7: Mesh Evaluation on ScanNet [9]. The estimated pose is shown in red, and the ground truth camera pose is shown in green. Our method stands out with its more accurate trajectory and higher quality reconstruction, such as the corners of the kitchen. Ours w/o strategic BA …
Figure 8
Figure 8. Figure 8: Mesh Evaluation on TUM RGB-D [54]. Our method stands out with its geometry details and higher texture fidelity. Without strategic BA (only with global BA), the performance can be suboptimal due to missing local information. rendering performance on the Replica dataset …
Figure 9
Figure 9. Figure 9: Ablation on loss term reweighting Ablation of strategic BA. Tab. 7 shows localization accuracy and rendering quality under different BA strategies on 6 ScanNet scenes. Experimental results demonstrate that our uncertainty-guided strategic BA method achieves optimal per…
Figure 10
Figure 10. Figure 10: Impact of Translational and Angular Velocities on Uncertainty. We can observe the correlation between uncertainty and both translational velocity and angular velocity. Higher velocities or accelerations tend to result in higher uncertainty. accelerations can easily ca…
Figure 11
Figure 11. Figure 11: Impact of different keyframe selection on Uncertainty. Here we compare the changing image-level uncertainty per frame with different keyframe selection strategies. The results indicated by the blue line show that image-level uncertainty is significantly reduced, achie…
Figure 12
Figure 12. Figure 12: Impact of Strategic BA on Rendering and Uncertainty Visualization. Our proposed strategic BA integrates global BA, local BA, and LLCO. This approach achieves the highest rendered image quality, as indicated by the PSNR (dB) metric. The second row presents visualized p…
Figure 13
Figure 13. Figure 13: Ablation on Gaussian Assumption Uncertainty Model. We use three grids to represent geometry, appearance, and learnable uncertainty respectively. variance β through a third grid parallel with the geometry and appearance grid. So, the term we need to minimize is: Lsingl…
Figure 15
Figure 15. Figure 15: Gaussian Assumption Model vs. Ours. Our model demonstrates superior rendering quality, as evaluated by PSNR (dB) ↑. Depth uncertainty, calculated using Eq. (1), is visualized for comparison. Our method visibly reduces depth uncertainty, as clearly shown in the visuali…
Figure 19
Figure 19. Figure 19: Comparative rendering quality convergence on TUM [PITH_FULL_IMAGE:figures/full_fig_p019_19.png]
Figure 18
Figure 18. Figure 18: Rendering Comparison on TUM RGB-D [54]. Ours shows the best results compared to state-of-the-art methods BSLAM [20] and Co-SLAM [62] among dense implicit SLAM methods. B.6. Runtime and Memory Analysis In Tab. 12, we compare runtime and memory usage, benchmarking all m…
Figure 20
Figure 20. Figure 20: Ablation on Reweighting. In the tracking and mapping processes, the loss functions consist of three loss terms: (Lsdf , Ldep, Lrgb). If reweighting is applied, we denote it as Y , and if not, we denote it as N. Column d) YYY-YYN indicates that we apply pixel-level unc…
Figure 21
Figure 21. Figure 21: Ablation on Model Design. We compare different scene representation combinations on Replica [53] room0 and evaluate with metrics PSNR and completion ratio[< 1cm%]. Hash-Hash-UncM denotes using hash grids for geometry and appearance, with a learnable uncertainty model.…
Figure 22
Figure 22. Figure 22: Mesh Evaluation on Replica [53] Office-0. Notably, our method can present fine geometric structures while also achieving better scene completion for unobserved regions compared to explicit Loopy-SLAM [30]. Compared to implicit methods such as Co-SLAM [62], BSLAM [20],…
Figure 23
Figure 23. Figure 23: Mesh Evaluation on Replica [53] Room-2. Our method achieves finer geometric and appearance reconstruction. For appearance: the patterns on the curtains and the detailed textures on the cabinet surface. For geometry: the vase on the cabinet and the cabinet legs. Please…
Figure 24
Figure 24. Figure 24: Mesh Evaluation on Replica [53] Office-2. For appearance: our rendered floor has higher quality, clearly distinguishing the floor patterns, as well as the textures on the pillows on the sofa. For geometry: we zoomed in on the table, and our method reconstructs sharper…
Figure 25
Figure 25. Figure 25: Mesh Evaluation on Replica [53] Office-4. For appearance: our rendered floor quality is higher, clearly distinguishing the floor patterns, as well as the clock on the wall. For geometry: we reconstructed more of the office chair’s geometric structure, such as the legs…
Figure 26
Figure 26. Figure 26: Mesh Evaluation on ScanNet [9]. Explicit Loopy-SLAM [30] and implicit Co-SLAM [62] are listed here for comparison. For appearance: Our method achieves higher rendering quality compared to the ground truth (GT) mesh, as seen texture of chairs, objects on the desk in sc…
Figure 27
Figure 27. Figure 27: Mesh Evaluation on TUM RGB-D [54]. Because there is no ground truth mesh for the TUM RGB-D dataset, we provide an image to facilitate qualitative comparison. We extensively compare the reconstruction quality with implicit methods such as ESLAM [22], Co-SLAM [62], and …
Figure 28
Figure 28. Figure 28: Mesh Evaluation on TUM RGB-D [54]. Because there is no ground truth mesh for the TUM RGB-D dataset, we provide an image to facilitate qualitative comparison. For example, our method accurately reconstructs details such as the Rubik’s cube on the table, the shopping ba…
Figure 29
Figure 29. Figure 29: Mesh Evaluation on TUM RGB-D [54]. While ESLAM [22] and BSLAM [20] can not capture geometric details such as cup and mouse on table, Co-SLAM [62] can not reconstruct thin geometric structure, such thin table surface. Our method shows outstanding performance. 20 [PITH…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. 3D Spatial Understanding in MLLMs: Disambiguation and Evaluation

    cs.CV 2024-12 conditional novelty 5.0 of 10

    A pipeline that adds explicit distractor and relative-position information to an MLLM improves generation of target-exclusive 3D referring instructions, validated partly by training 3D grounding models on the generated text.

Reference graph

Works this paper leans on

82 extracted references · 73 canonical work pages · cited by 1 Pith paper

  1. [1]

    Uncertainty in Contrastive Learning: On the Predictability of Downstream Performance

    Shervin Ardeshir and Navid Azizan. Uncertainty in contrastive learning: On the predictability of downstream performance. arXiv preprint arXiv:2207.09336, 2022. 2

    work page Pith review arXiv 2022

  2. [2]

    Neural rgb-d surface reconstruction

    Dejan Azinovi ´c, Ricardo Martin-Brualla, Dan B Goldman, Matthias Nießner, and Justus Thies. Neural rgb-d surface reconstruction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 6290–6301, 2022. 2, 6, 1

    2022

  3. [3]

    On the comparison of uncertainty criteria for active slam

    Henry Carrillo, Ian Reid, and Jos ´e A Castellanos. On the comparison of uncertainty criteria for active slam. In 2012 IEEE International Conference on Robotics and Automation, pages 2080–2087. IEEE, 2012. 2

    2012

  4. [4]

    Efficient geometry-aware 3d generative adversarial networks

    Eric R Chan, Connor Z Lin, Matthew A Chan, Koki Nagano, Boxiao Pan, Shalini De Mello, Orazio Gallo, Leonidas J Guibas, Jonathan Tremblay, Sameh Khamis, et al. Efficient geometry-aware 3d generative adversarial networks. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 16123–16133, 2022. 2, 8, 6

    work page 2022

  5. [5]

    Pixel-wise prediction based visual odometry via uncertainty estimation

    Hao-Wei Chen, Ting-Hsuan Liao, Hsuan-Kung Yang, and Chun-Yi Lee. Pixel-wise prediction based visual odometry via uncertainty estimation. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , pages 2518–2528, 2023. 2

    work page 2023

  6. [6]

    Sp-slam: Surfel-point simultaneous localization and mapping

    Hae Min Cho, HyungGi Jo, and Euntai Kim. Sp-slam: Surfel-point simultaneous localization and mapping. IEEE/ASME Transactions on Mechatronics , 27(5):2568–2579, 2021. 2

    work page 2021

  7. [7]

    Orbeez-slam: A real-time monocular visual slam with orb features and nerf-realized mapping

    Chi-Ming Chung, Yang-Che Tseng, Ya-Ching Hsu, Xiang-Qian Shi, Yun-Hung Hua, Jia-Fong Yeh, Wen-Chin Chen, Yi-Ting Chen, and Winston H Hsu. Orbeez-slam: A real-time monocular visual slam with orb features and nerf-realized mapping. In 2023 IEEE International Conference on Robotics and Automation (ICRA) , pages 9400–9406. IEEE, 2023. 2

    work page 2023

  8. [8]

    A volumetric method for building complex models from range images

    Brian Curless and Marc Levoy. A volumetric method for building complex models from range images. InProceedings of the 23rd annual conference on Computer graphics and interactive techniques, pages 303–312, 1996. 2, 7

    work page 1996

Show all 82 references
  1. [9]

    Scannet: Richly-annotated 3d reconstructions of indoor scenes

    Angela Dai, Angel X Chang, Manolis Savva, Maciej Halber, Thomas Funkhouser, and Matthias Nießner. Scannet: Richly-annotated 3d reconstructions of indoor scenes. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 5828–5839, 2017. 2, 6, 7, 1,...

    2017

  2. [10]

    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. 8

    2017

  3. [11]

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

    Tianchen Deng, Guole Shen, Tong Qin, Jianyu Wang, Wentao 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 Recogn...

    2024

  4. [12]

    Lsd-slam: Large-scale direct monocular slam

    Jakob Engel, Thomas Sch ¨ops, and Daniel Cremers. Lsd-slam: Large-scale direct monocular slam. In European conference on computer vision , pages 834–849. Springer,

  5. [13]

    Naruto: Neural active reconstruction from uncertain target observations

    Ziyue Feng, Huangying Zhan, Zheng Chen, Qingan Yan, Xiangyu Xu, Changjiang Cai, Bing Li, Qilun Zhu, and Yi Xu. Naruto: Neural active reconstruction from uncertain target observations. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages ...

    2024

  6. [14]

    Svo: Fast semi-direct monocular visual odometry

    Christian Forster, Matia Pizzoli, and Davide Scaramuzza. Svo: Fast semi-direct monocular visual odometry. In 2014 IEEE international conference on robotics and automation (ICRA), pages 15–22. IEEE, 2014. 1

    2014

  7. [15]

    Bayes’ rays: Uncertainty quantification for neural radiance fields

    Lily Goli, Cody Reading, Silvia Sell ´an, Alec Jacobson, and Andrea Tagliasacchi. Bayes’ rays: Uncertainty quantification for neural radiance fields. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20061–20070, 2024. 2

    2024

  8. [16]

    Bayes’ rays: Uncertainty quantification for neural radiance fields

    Lily Goli, Cody Reading, Silvia Sell ´an, Alec Jacobson, and Andrea Tagliasacchi. Bayes’ rays: Uncertainty quantification for neural radiance fields. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20061–20070, 2024. 4

    2024

  9. [17]

    Efficient uncertainty estimation in semantic segmentation via distillation

    Christopher J Holder and Muhammad Shafique. Efficient uncertainty estimation in semantic segmentation via distillation. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 3087–3094, 2021. 2

    2021

  10. [18]

    Gradient-based uncertainty for monocular depth estimation

    Julia Hornauer and Vasileios Belagiannis. Gradient-based uncertainty for monocular depth estimation. In European Conference on Computer Vision , pages 613–630. Springer,

  11. [19]

    Out-of-distribution detection for monocular depth estimation

    Julia Hornauer, Adrian Holzbock, and Vasileios Belagiannis. Out-of-distribution detection for monocular depth estimation. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 1911–1921, 2023. 2

    1911

  12. [20]

    Benchmarking implicit neural representation and geometric rendering in real-time rgb-d slam

    Tongyan Hua and Lin Wang. Benchmarking implicit neural representation and geometric rendering in real-time rgb-d slam. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 21346–21356, 2024. 1, 2, 5, 6, 7, 8, 4, 11, 12, 13, 18, 20

    2024

  13. [21]

    Aleatoric and epistemic uncertainty in machine learning: An introduction to concepts and methods

    Eyke H ¨ullermeier and Willem Waegeman. Aleatoric and epistemic uncertainty in machine learning: An introduction to concepts and methods. Machine learning, 110(3):457–506, 2021. 2

    2021

  14. [22]

    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. 1, 2, 6, 7,...

    2023

  15. [23]

    Values: A framework for systematic validation of uncertainty estimation in semantic segmentation

    Kim-Celine Kahl, Carsten T L ¨uth, Maximilian Zenk, Klaus Maier-Hein, and Paul F Jaeger. Values: A framework for systematic validation of uncertainty estimation in semantic segmentation. arXiv preprint arXiv:2401.08501, 2024. 2

    2024 arXiv

  16. [24]

    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 9 Jonathon Luiten. Splatam: Splat track & map 3d gaussians for dense rgb-d slam. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition ...

    2024

  17. [25]

    What uncertainties do we need in bayesian deep learning for computer vision? Advances in neural information processing systems, 30, 2017

    Alex Kendall and Yarin Gal. What uncertainties do we need in bayesian deep learning for computer vision? Advances in neural information processing systems, 30, 2017. 2

    2017

  18. [26]

    Parallel tracking and mapping for small ar workspaces

    Georg Klein and David Murray. Parallel tracking and mapping for small ar workspaces. In 2007 6th IEEE and ACM international symposium on mixed and augmented reality, pages 225–234. IEEE, 2007. 2

    2007

  19. [27]

    vmap: Vectorised object mapping for neural field slam

    Xin Kong, Shikun Liu, Marwan Taher, and Andrew J Davison. vmap: Vectorised object mapping for neural field slam. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 952–961,

  20. [28]

    Uncertainty guided policy for active robotic 3d reconstruction using neural radiance fields

    Soomin Lee, Le Chen, Jiahao Wang, Alexander Liniger, Suryansh Kumar, and Fisher Yu. Uncertainty guided policy for active robotic 3d reconstruction using neural radiance fields. IEEE Robotics and Automation Letters , 7(4):12070–12077, 2022. 2

    2022

  21. [29]

    Deepslam: A robust monocular slam system with unsupervised deep learning

    Ruihao Li, Sen Wang, and Dongbing Gu. Deepslam: A robust monocular slam system with unsupervised deep learning. IEEE Transactions on Industrial Electronics , 68(4):3577–3587, 2020. 2

    2020

  22. [30]

    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, 8, 13, 17, 18

    2024

  23. [31]

    Marching cubes: A high resolution 3d surface construction algorithm

    William E Lorensen and Harvey E Cline. Marching cubes: A high resolution 3d surface construction algorithm. ACM SIGGRAPH Computer Graphics, 21(4):163–169, 1987. 6, 1

    1987

  24. [32]

    Uncertainty-weighted loss functions for improved adversarial attacks on semantic segmentation

    Kira Maag and Asja Fischer. Uncertainty-weighted loss functions for improved adversarial attacks on semantic segmentation. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , pages 3906–3914, 2024. 2

    2024

  25. [33]

    Ngel-slam: Neural implicit representation-based global consistent low-latency slam system

    Yunxuan Mao, Xuan Yu, Kai Wang, Yue Wang, Rong Xiong, and Yiyi Liao. Ngel-slam: Neural implicit representation-based global consistent low-latency slam system. arXiv preprint arXiv:2311.09525, 2023. 2

    2023 arXiv

  26. [34]

    Nerf in the wild: Neural radiance fields for unconstrained photo collections

    Ricardo Martin-Brualla, Noha Radwan, Mehdi SM Sajjadi, Jonathan T Barron, Alexey Dosovitskiy, and Daniel Duckworth. Nerf in the wild: Neural radiance fields for unconstrained photo collections. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognitio...

    2021

  27. [35]

    Gaussian splatting slam

    Hidenobu Matsuki, Riku Murai, Paul HJ Kelly, and Andrew J Davison. Gaussian splatting slam. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18039–18048, 2024. 2

    2024

  28. [36]

    Nerf: Representing scenes as neural radiance fields for view synthesis

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view synthesis. Communications of the ACM , 65(1):99–106,

  29. [37]

    Instant neural graphics primitives with a multiresolution hash encoding

    Thomas M ¨uller, Alex Evans, Christoph Schied, and Alexander Keller. Instant neural graphics primitives with a multiresolution hash encoding. ACM Transactions on Graphics (ToG), 41(4):1–15, 2022. 2, 3, 8

    2022

  30. [38]

    Orb-slam2: An open-source slam system for monocular, stereo, and rgb-d cameras

    Raul 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,

  31. [39]

    Kinectfusion: Real-time dense surface mapping and tracking

    Richard A Newcombe, Shahram Izadi, Otmar Hilliges, David Molyneaux, David Kim, Andrew J Davison, Pushmeet Kohi, Jamie Shotton, Steve Hodges, and Andrew Fitzgibbon. Kinectfusion: Real-time dense surface mapping and tracking. In 2011 10th IEEE international symposium on mixed an...

    2011

  32. [40]

    Dtam: Dense tracking and mapping in real-time

    Richard A Newcombe, Steven J Lovegrove, and Andrew J Davison. Dtam: Dense tracking and mapping in real-time. In 2011 international conference on computer vision , pages 2320–2327. IEEE, 2011. 1, 2

    2011

  33. [41]

    Real-time 3d reconstruction at scale using voxel hashing

    Matthias Nießner, Michael Zollh ¨ofer, Shahram Izadi, and Marc Stamminger. Real-time 3d reconstruction at scale using voxel hashing. ACM Transactions on Graphics (ToG), 32(6):1–11, 2013. 2

    2013

  34. [42]

    Stylesdf: High-resolution 3d-consistent image and geometry generation

    Roy Or-El, Xuan Luo, Mengyi Shan, Eli Shechtman, Jeong Joon Park, and Ira Kemelmacher-Shlizerman. Stylesdf: High-resolution 3d-consistent image and geometry generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 13503–13513, 2022. 3

    2022

  35. [43]

    Activenerf: Learning where to see with uncertainty estimation

    Xuran Pan, Zihang Lai, Shiji Song, and Gao Huang. Activenerf: Learning where to see with uncertainty estimation. In European Conference on Computer Vision , pages 230–246. Springer, 2022. 2

    2022

  36. [44]

    Nerf on-the-go: Exploiting uncertainty for distractor-free nerfs in the wild

    Weining Ren, Zihan Zhu, Boyang Sun, Jiaqi Chen, Marc Pollefeys, and Songyou Peng. Nerf on-the-go: Exploiting uncertainty for distractor-free nerfs in the wild. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8931–8940, 2024. 5

    2024

  37. [45]

    Nerf-slam: Real-time dense monocular slam with neural radiance fields

    Antoni Rosinol, John J Leonard, and Luca Carlone. Nerf-slam: Real-time dense monocular slam with neural radiance fields. In 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 3437–3444. IEEE, 2023. 2

    2023

  38. [46]

    Probabilistic volumetric fusion for dense monocular slam

    Antoni Rosinol, John J Leonard, and Luca Carlone. Probabilistic volumetric fusion for dense monocular slam. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , pages 3097–3105, 2023. 2

    2023

  39. [47]

    Point-slam: Dense neural point cloud-based slam

    Erik Sandstr ¨om, Yue Li, Luc Van Gool, and Martin R Oswald. Point-slam: Dense neural point cloud-based slam. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 18433–18444, 2023. 2, 8, 7

    2023

  40. [48]

    Uncle-slam: Uncertainty learning for dense neural slam

    Erik Sandstr ¨om, Kevin Ta, Luc Van Gool, and Martin R Oswald. Uncle-slam: Uncertainty learning for dense neural slam. In Proceedings of the International Conference on Computer Vision (ICCV) Workshops. IEEE, 2023. Available as arXiv preprint arXiv:2306.11048. 2

    2023 arXiv

  41. [49]

    Deeptio: A deep thermal-inertial odometry with visual hallucination

    Muhamad Risqi U Saputra, Pedro PB De Gusmao, Chris Xiaoxuan Lu, Yasin Almalioglu, Stefano Rosa, 10 Changhao Chen, Johan Wahlstr ¨om, Wei Wang, Andrew Markham, and Niki Trigoni. Deeptio: A deep thermal-inertial odometry with visual hallucination. IEEE Robotics and Automation Le...

    2020

  42. [50]

    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. 8

    2019

  43. [51]

    Conditional-flow nerf: Accurate 3d modelling with reliable uncertainty quantification

    Jianxiong Shen, Antonio Agudo, Francesc Moreno-Noguer, and Adria Ruiz. Conditional-flow nerf: Accurate 3d modelling with reliable uncertainty quantification. In European Conference on Computer Vision, pages 540–557. Springer, 2022. 2

    2022

  44. [52]

    Stochastic neural radiance fields: Quantifying uncertainty in implicit 3d representations

    Jianxiong Shen, Adria Ruiz, Antonio Agudo, and Francesc Moreno-Noguer. Stochastic neural radiance fields: Quantifying uncertainty in implicit 3d representations. In 2021 International Conference on 3D Vision (3DV) , pages 972–981. IEEE, 2021. 2

    2021

  45. [53]

    The replica dataset: A digital replica of indoor spaces

    Julian Straub, Thomas Whelan, Lingni Ma, Yufan Chen, Erik Wijmans, Simon Green, Jakob J Engel, Raul Mur-Artal, Carl Ren, Shobhit Verma, et al. The replica dataset: A digital replica of indoor spaces. arXiv preprint arXiv:1906.05797,

    1906 arXiv

  46. [54]

    A benchmark for the evaluation of rgb-d slam systems

    J ¨urgen Sturm, Nikolas Engelhard, Felix Endres, Wolfram Burgard, and Daniel Cremers. A benchmark for the evaluation of rgb-d slam systems. In 2012 IEEE/RSJ international conference on intelligent robots and systems , pages 573–580. IEEE, 2012. 1, 2, 6, 7, 8, 5, 18, 19, 20

    2012

  47. [55]

    A benchmark for the evaluation of rgb-d slam systems

    J ¨urgen Sturm, Nikolas Engelhard, Felix Endres, Wolfram Burgard, and Daniel Cremers. A benchmark for the evaluation of rgb-d slam systems. In 2012 IEEE/RSJ international conference on intelligent robots and systems , pages 573–580. IEEE, 2012. 6

    2012

  48. [56]

    imap: Implicit mapping and positioning in real-time

    Edgar Sucar, Shikun Liu, Joseph Ortiz, and Andrew J Davison. imap: Implicit mapping and positioning in real-time. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 6229–6238, 2021. 1, 2, 5, 6, 7, 8

    2021

  49. [57]

    Neuralrecon: Real-time coherent 3d reconstruction from monocular video

    Jiaming Sun, Yiming Xie, Linghao Chen, Xiaowei Zhou, and Hujun Bao. Neuralrecon: Real-time coherent 3d reconstruction from monocular video. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 15598–15607, 2021. 2

    2021

  50. [58]

    Density-aware nerf ensembles: Quantifying predictive uncertainty in neural radiance fields

    Niko S ¨underhauf, Jad Abou-Chakra, and Dimity Miller. Density-aware nerf ensembles: Quantifying predictive uncertainty in neural radiance fields. In 2023 IEEE International Conference on Robotics and Automation (ICRA), pages 9370–9376. IEEE, 2023. 2, 4

    2023

  51. [59]

    Neural geometric level of detail: Real-time rendering with implicit 3d shapes

    Towaki Takikawa, Joey Litalien, Kangxue Yin, Karsten Kreis, Charles Loop, Derek Nowrouzezahrai, Alec Jacobson, Morgan McGuire, and Sanja Fidler. Neural geometric level of detail: Real-time rendering with implicit 3d shapes. In Proceedings of the IEEE/CVF Conference on Computer...

    2021

  52. [60]

    Mips-fusion: Multi-implicit-submaps for scalable and robust online neural rgb-d reconstruction.ACM Transactions on Graphics (TOG), 42(6):1–16, 2023

    Yijie Tang, Jiazhao Zhang, Zhinan Yu, He Wang, and Kai Xu. Mips-fusion: Multi-implicit-submaps for scalable and robust online neural rgb-d reconstruction.ACM Transactions on Graphics (TOG), 42(6):1–16, 2023. 2, 6, 7, 8

    2023

  53. [61]

    Droid-slam: Deep visual slam for monocular, stereo, and rgb-d cameras

    Zachary Teed and Jia Deng. Droid-slam: Deep visual slam for monocular, stereo, and rgb-d cameras. Advances in neural information processing systems, 34:16558–16569,

  54. [62]

    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 Conference on Computer Vision and Pattern Recognition , pages 13293–13302, 2023. 1, 2, 3, 4, 5, 6, 7, 8, 11, 12, ...

    2023

  55. [63]

    Neus: Learning neural implicit surfaces by volume rendering for multi-view reconstruction

    Peng Wang, Lingjie Liu, Yuan Liu, Christian Theobalt, Taku Komura, and Wenping Wang. Neus: Learning neural implicit surfaces by volume rendering for multi-view reconstruction. Advances in neural information processing systems, 2021. 3

    2021

  56. [64]

    Image quality assessment: from error visibility to structural similarity

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Simoncelli. Image quality assessment: from error visibility to structural similarity. IEEE transactions on image processing, 13(4):600–612, 2004. 6

    2004

  57. [65]

    Elasticfusion: Dense slam without a pose graph

    Thomas Whelan, Stefan Leutenegger, Renato Salas-Moreno, Ben Glocker, and Andrew Davison. Elasticfusion: Dense slam without a pose graph. In Proceedings of the Robotics: Science and Systems. RSS, 2015. 2

    2015

  58. [66]

    Comparison of uncertainty quantification methods for cnn-based regression

    Kira Wursthorn, Markus Hillemann, and Markus Ulrich. Comparison of uncertainty quantification methods for cnn-based regression. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 43:721–728, 2022. 2

    2022

  59. [67]

    Debsdf: Delving into the details and bias of neural indoor scene reconstruction

    Yuting Xiao, Jingwei Xu, Zehao Yu, and Shenghua Gao. Debsdf: Delving into the details and bias of neural indoor scene reconstruction. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024. 5

    2024

  60. [68]

    Point-nerf: Point-based neural radiance fields

    Qiangeng Xu, Zexiang Xu, Julien Philip, Sai Bi, Zhixin Shu, Kalyan Sunkavalli, and Ulrich Neumann. Point-nerf: Point-based neural radiance fields. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5438–5448, 2022. 2

    2022

  61. [69]

    Neural visibility field for uncertainty-driven active mapping

    Shangjie Xue, Jesse Dill, Pranay Mathur, Frank Dellaert, Panagiotis Tsiotra, and Danfei Xu. Neural visibility field for uncertainty-driven active mapping. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18122–18132, 2024. 2

    2024

  62. [70]

    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 Conference on Computer Vision and Pattern Recognition , pages 19595–19604, 2024. 2

    2024

  63. [71]

    V ox-fusion: Dense tracking and mapping with voxel-based neural implicit representation

    Xingrui Yang, Hai Li, Hongjia Zhai, Yuhang Ming, Yuqian Liu, and Guofeng Zhang. V ox-fusion: Dense tracking and mapping with voxel-based neural implicit representation. In 2022 IEEE International Symposium on Mixed and Augmented Reality (ISMAR) , pages 499–507. IEEE, 2022. 2, 3, 12

    2022

  64. [72]

    V olume rendering of neural implicit surfaces

    Lior Yariv, Jiatao Gu, Yoni Kasten, and Yaron Lipman. V olume rendering of neural implicit surfaces. Advances in Neural Information Processing Systems , 34:4805–4815,

  65. [73]

    Vitamin-e: Visual tracking and mapping with extremely dense feature points

    Masashi Yokozuka, Shuji Oishi, Simon Thompson, and Atsuhiko Banno. Vitamin-e: Visual tracking and mapping with extremely dense feature points. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 9641–9650, 2019. 2

    2019

  66. [74]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 586–595, 2018. 6

    2018

  67. [75]

    Go-slam: Global optimization for consistent 3d instant reconstruction

    Youmin Zhang, Fabio Tosi, Stefano Mattoccia, and Matteo Poggi. Go-slam: Global optimization for consistent 3d instant reconstruction. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 3727–3737, 2023. 2, 3

    2023

  68. [76]

    Nicer-slam: Neural implicit scene encoding for rgb slam

    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. In 2024 International Conference on 3D Vision (3DV) , pages 42–52. IEEE, 2024. 2, 5

    2024

  69. [77]

    Nice-slam: Neural implicit scalable encoding for slam

    Zihan Zhu, Songyou Peng, Viktor Larsson, Weiwei Xu, Hujun Bao, Zhaopeng Cui, Martin R Oswald, and Marc Pollefeys. Nice-slam: Neural implicit scalable encoding for slam. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 12786–12796, 20...

    2022

  70. [79]

    The learning rate of translation is set to 5e−4, and the learning rate of rotation is 3e−3

    Because of invalid depth at the edge of the image of ScanNet, 75 pixels are culled at the edge of the image for tracking during data pre-processing. The learning rate of translation is set to 5e−4, and the learning rate of rotation is 3e−3. TUM RGB-D Dataset [54] The image-lev...

  71. [80]

    However, in unobserved regions where the density values σ(pk) are very small or zero, the exponential term will tend to 1, so NX n=1 wn = 1 − 1 = 0 Therefore, the termination probability is proven to lie within the range (0, 1). A.3. Co-visibility Check Loop detection is imple...

    2000

  72. [81]

    3.4 in the main paper)

    (Sec. 3.4 in the main paper). The experiment is conducted on Replica [53], and the metrics are ATE RMSE (cm), reconstruction accuracy (cm), reconstruction completion (cm), completion ratio and PSNR. BSLAM [20] can also benefit from our strategy. Method ATE Acc. Comp. Ratio PSN...

  73. [82]

    method for encoding geometry and appearance respectively. In Tab. 13, rows a) through d) provide quantitative results on the Replica dataset, while Fig. 21 presents the corresponding qualitative visualizations. The results show that using two hash grids for encoding provides t...

  74. [2019]

    2, 6, 7, 8, 1, 4, 5, 10, 11, 12, 13, 14, 15, 16

Pith tools

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