Pith. sign in

REVIEW 3 major objections 6 minor 5 cited by

PreF3R: Pose-Free Feed-Forward 3D Gaussian Splatting from Variable-length Image Sequence

T0 review · 3 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read PreF3R reconstructs a 3D Gaussian scene from a variable-length unposed image sequence in a single feed-forward pass at 20 FPS.

desk verdict PreF3R is a credible pose-free feed-forward NVS pipeline for variable-length sequences, but the 'unlimited length' claim needs to be dialed back; the 50-view drift and memory-eviction mechanism don't support it. read the letter →

arxiv 2411.16877 v1 pith:PVPLR5MG submitted 2024-11-25 cs.CV

Zequn Chen , Jiezhi Yang , Heng Yang This is my paper · ORCID
classification cs.CV
keywords pose-free3Dreconstructionfeed-forwardGaussiansplattingvariable-lengthimagesequencespatialmemorynetworknovel-viewsynthesisDUSt3Rprimitives
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

PreF3R claims to be the first system that turns an ordered sequence of unposed images, of any length, directly into a set of 3D Gaussian primitives in a single feed-forward pass, with no camera calibration, no per-scene optimization, and no global alignment step. If true, a robot or AR device could stream video frames and obtain a renderable 3D model of the scene in real time, at 20 FPS reconstruction and 200 FPS novel-view rendering. The paper builds on an existing pairwise 3D reconstruction model and a spatial memory mechanism that anchors all predicted pointmaps to the coordinate frame of the first frame, then adds a head that predicts Gaussian parameters so the result can be supervised by both geometric and photometric losses.

What carries the argument

The load-bearing mechanism is the spatial memory network: a set of memory banks that store key and value features from previous frames, queried by a cross-attention decoder so that each new image is registered and fused into the coordinate frame of the first input view. This is what removes the optimization-based global alignment step and lets the sequence length be arbitrary; the working-memory and long-term-memory split keeps GPU usage roughly constant regardless of sequence length. The second piece is the Gaussian parameter head, a dense MLP parallel to the pointmap head that predicts per-pixel opacity, covariance, and spherical-harmonic color, enabling differentiable rasterization and photometric supervision.

What would settle it

Run PreF3R on a long closed-loop video, for instance 500 frames that revisit the starting viewpoint, and measure whether the reconstructed Gaussians of the revisited region reproject consistently; if reprojection error grows with loop length, or if PSNR on a fixed evaluation view drops with sequence length (the paper's own tables show PSNR falling from 22.60 at 10 views to 20.38 at 50 views on ScanNet++), then the canonical-frame stability assumption is falsified.

Watch

Extended reading notes

Core claim

The central claim is that 3D Gaussian reconstruction and novel-view synthesis can be done feed-forward, pose-free, and online from a variable-length sequence. The model takes frames one by one; each new frame is encoded, matched against a memory of previous frames, and decoded into both a pointmap and a set of pixel-aligned Gaussian parameters in a common canonical space, namely the first frame's coordinate system. Because the memory network maintains this common space incrementally, there is no need for the global alignment optimization that pairwise reconstruction methods would require for multi-view inputs. The predicted Gaussians are rasterized with the standard differentiable Gaussian splatting renderer, and training jointly minimizes a confidence-weighted pointmap regression loss and a masked photometric loss against ground-truth depth and RGB. The paper reports 20 FPS reconstruction and 200 FPS rendering on an H100 GPU, with competitive or better PSNR, SSIM, and LPIPS than pose-free and pose-based baselines on ScanNet++ and ARKitScenes.

Load-bearing premise

The spatial memory network must keep the canonical 3D frame stable and correctly scaled over an arbitrarily long sequence without any global optimization; if drift accumulates, all Gaussians after the drift point are misaligned and rendering quality degrades.

Editorial extensions

If this is right

  • A camera rig or phone can sweep through a scene and obtain a renderable 3D Gaussian model at 20 FPS, with no prior pose estimation, enabling real-time AR, VR, and robotics mapping pipelines.
  • Sequence length can be arbitrary: because memory banks are sparsified and bounded, GPU memory stays roughly flat as more frames arrive.
  • Because the same architecture produces both pointmaps and Gaussian parameters, it can be trained with both geometric pointmap regression and photometric rendered-image losses, improving structural and visual accuracy; the paper's ablation shows the masked photometric loss is critical.
  • The model generalizes to unseen scenes without finetuning: cross-dataset evaluations on ScanNet++ and ARKitScenes still beat pose-free baselines at 10 views, and it also runs on out-of-domain Tanks and Temples scenes.

Reading between the lines

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

  • Beyond the paper's explicit claims: because the memory is anchored to the first frame, the approach behaves as a monocular SLAM-style odometry in Gaussian space; a loop-closure or global bundle adjustment would likely be needed to keep long trajectories drift-free.
  • The same architecture could potentially be trained with self-supervised photometric loss alone (dropping depth supervision) if scale ambiguity is handled, opening the door to training on internet-scale unposed video.
  • The confidence-based Gaussian pruning suggests a natural way to filter floaters and spurious geometry that could transfer to other feed-forward Gaussian prediction models.
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

3 major / 6 minor

Summary. The paper presents PreF3R, a pose-free, feed-forward framework that reconstructs a global 3D Gaussian field from a variable-length sequence of unposed images. It builds on DUSt3R's pairwise pointmap prediction, extends it to arbitrary-length input via a spatial memory network (adopted from Spann3R), and adds a dense Gaussian parameter head so that a differentiable rasterizer can render novel views. The model is trained with a combination of confidence-weighted pointmap regression and masked photometric losses, and is evaluated on ScanNet++ and ARKitScenes at 2, 10, and 50 input views, with comparisons to MVSplat, InstantSplat, Spann3R, and Splatt3R. The authors report 20 FPS incremental reconstruction and 200 FPS novel-view rendering without per-scene optimization, and claim to be the first pose-free feed-forward Gaussian reconstruction method that handles variable-length sequences.

Significance. If the claims hold, this is a useful step toward practical online 3D reconstruction from unposed video: it removes both camera calibration and per-scene optimization, and it demonstrates that a pretrained DUSt3R-style backbone can be extended with a memory mechanism and a Gaussian head to produce renderable 3D representations in a single feed-forward pass. The paper includes a reasonable ablation study (Table 3) showing that each component contributes, and it reports cross-dataset evaluations in the supplementary material. The main weaknesses are that the 'unlimited-length' claim is not supported by the evidence (the model is trained on 5 views, and performance degrades markedly at 50 views), and that the evaluation is based on only 10 validation scenes per dataset without error bars or geometric metrics. These issues are load-bearing for the paper's central generalization claim.

major comments (3)
  1. [Sec. 1, Sec. 3.2, Tabs. 1-2] The paper's central claim is that PreF3R generalizes to 'an ordered image sequence of unlimited length' (Sec. 1) and maintains a globally consistent Gaussian field in the first view's coordinate system. This claim is not supported by the evidence. The spatial memory in Sec. 3.2 is bounded: working memory keeps only the Nworking most recent frames, and long-term memory retains only the top-k tokens per token based on accumulated attention. Once early frames, including the anchor frame, are sparsified or evicted, no mechanism enforces alignment with the first view's frame. Training uses Ntrain=5 views (Sec. 4.1), so the 10- and 50-view evaluations are extrapolations beyond the training regime. The paper's own results show monotonic degradation as the sequence length grows: PSNR on ScanNet++ falls from 22.60 (10 views) to 20.38 (50 views) in Tab. 1, and on ARKitScenes from 21.91 to 18.70 in Tab. 2. The only reported metrics are novel-view rendering; no metric directly measures canonical-frame drift or geometric consistency, so it is impossible to tell whether the degradation is due to accumulated drift, memory sparsification, or simply the difficulty of longer sequences. The limitation statement in Sec. 4.4 that a low-overlap transition degrades 'all subsequent frames following It+1' confirms that unbounded error propagation is a risk. I recommend either tempering the 'unlimited length' claim to 'sequences substantially longer than the training length, with degradation characterized,' or adding direct drift measurements (e.g., alignment error between predicted pointmaps and ground-truth geometry in the first-view frame) and evaluating at T=100 and T=200.
  2. [Sec. 4.2, Tabs. 1-2] The evaluation is statistically thin. All results are averaged over only 10 validation scenes per dataset, with no per-scene breakdown, standard deviation, or confidence intervals. Given that the differences between methods are often small (e.g., Tab. 1 at 10 views: ours 22.60 vs Spann3R 21.86, a 0.74 dB gap), it is unclear whether the reported improvements are significant. The 50-view degradation is likewise reported as a single mean without variance. Furthermore, the paper is about 3D reconstruction, yet no geometric metric (e.g., pointmap chamfer distance, F-score, or alignment error) is reported. The pointmap regression loss Lconf is part of training, but the predicted pointmaps in the canonical frame are never evaluated against ground truth. Adding geometric metrics would directly test the canonical-frame consistency claim and would make the 'robust generalization' conclusion more credible.
  3. [Sec. 4.2, Tabs. 1-2, Supp. Fig. 6] The comparison with Spann3R is not a like-for-like rendering evaluation. Spann3R predicts colored pointmaps, and the authors evaluate it by projecting those pointmaps back onto image planes. As the authors themselves note in the supplementary material (Fig. 6), this process can produce black areas and floaters, which disadvantages Spann3R relative to a method with a proper Gaussian rasterizer. The performance gap over Spann3R may therefore be largely a rendering-representation artifact rather than a reconstruction-quality improvement. A fairer baseline would be to attach a Gaussian head to Spann3R (using the same rasterizer) or to compare against other feed-forward multi-view Gaussian methods trained on the same data. This is important because the paper's headline 'outperforms existing baselines' depends on this comparison.
minor comments (6)
  1. [References, Sec. 4.2] The bibliography entry [32] is for 'Splatam: Splat track & map 3d gaussians for dense rgb-d slam,' but the text cites [32] for Splatt3R, which is reference [51] in the bibliography. Please correct the citation of Splatt3R throughout Sec. 4.2 and Tabs. 1-2.
  2. [Sec. 3.3, Eq. (5)] Eq. (5) uses f_h^t for the Gaussian head output, but Sec. 3.2 defines the reference decoder output as f_h^{t-1}; please make the index notation consistent.
  3. [Sec. 3.4, Eq. (7)] Eq. (7) appears to have a garbled rendering in the manuscript, with the norm notation missing from the regression loss. Please ensure the formula is typeset correctly.
  4. [Sec. 3.4, Eq. (8)] The mask in Eq. (8) is defined only by text ('setting the region where predicted alpha is less than th_alpha to zero'); please specify the element-wise multiplication explicitly.
  5. [Sec. 4.2] In Sec. 4.2, 'IntantSplat' is a typo for 'InstantSplat.' Also, capitalization of 'ScanNet++' and 'ARKitScenes' should be made consistent throughout the text.
  6. [Sec. 1, Sec. 2.3] The claim of being 'the first pose-free, feed-forward framework for online 3D Gaussian reconstruction from a variable-length image sequence' should be qualified with a discussion of how it relates to Spann3R, which already provides variable-length pose-free pointmap reconstruction, and to Splatt3R, which provides pose-free Gaussians for pairs.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: PreF3R is trained and evaluated on held-out scenes with standard supervised losses; adopting DUSt3R/Spann3R as pretrained building blocks is legitimate reuse, not a self-referential derivation.

full rationale

The paper's derivation chain is not circular. PreF3R builds on pretrained DUSt3R-style structural reconstruction and a Spann3R-style spatial memory network, but these are external prior works, not the authors' own unverified premises. The paper's actual contributions—the Gaussian parameter prediction head, joint photometric and pointmap supervision, and incremental feed-forward fusion—are trained on official training splits of ScanNet, ScanNet++, and ARKitScenes and evaluated on separate validation scenes (Sec. 4.1). No parameter is fitted to the evaluation targets; rendering PSNR/SSIM/LPIPS are computed from held-out novel views using dataset ground-truth poses, so the reported numbers are genuine predictions rather than identities. The scale factor z in Eq. 7 is a standard normalization for scale ambiguity, not a fitted parameter renamed as a prediction. The only self-citations (Refs. [69], [76]) appear in related-work context and are not load-bearing. The paper's explicitly stated limitations—degradation after low-overlap transitions, single-room indoor bias, 224x224 resolution, and the 50-view PSNR drop in Tables 1-2—are generalization risks, not evidence that evaluation targets were baked into the model. This is a standard build-on-prior-work paper, and no equation reduces to its own input.

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

The ledger shows one hand-tuned hyperparameter set and several domain assumptions about scale reliability, memory consistency, and dataset transferability. No new physical or mathematical entities are introduced. The method's contribution is architectural combination and training, not a new theoretical principle.

free parameters (5)
  • lambda (photometric loss weight) = 0.1
    Chosen by hand in Eq. (9) to balance confidence-based pointmap loss and masked MSE rendering loss; affects reconstruction versus photorealism trade-off.
  • alpha (confidence regularization) = 0.4
    Regularization hyperparameter in the confidence loss Eq. (6), chosen by hand following DUSt3R practice.
  • th_alpha (photometric loss mask threshold) = 1e-3
    Threshold on predicted alpha for masking the photometric loss in Eq. (8); chosen by hand, ablation shows it matters.
  • th_conf (inference confidence pruning threshold) = 1.0
    Confidence threshold for pruning predicted Gaussians at inference; chosen by hand, ablation shows it improves quality.
  • Ntrain, Nextra, Tmin, Tmax (training sampling parameters) = 5, 2, 5, 10
    Number of input views, extra photometric views, and frame sampling interval; chosen by hand to balance memory and overlap.
assumptions (4)
  • domain assumption Training on ScanNet, ScanNet++, and ARKitScenes with metric depth and ground-truth poses transfers to other indoor scenes.
    The paper evaluates only on validation splits of the same two indoor datasets (plus a qualitative Tanks and Temples figure), so cross-domain generalization is asserted rather than demonstrated.
  • domain assumption The pointmap scale factor z computed from the predicted global pointmap norm is reliable on the chosen training datasets.
    Sec. 3.4 explicitly states 'we choose datasets where the scale factor is sufficiently well estimated to train our model,' which is an acknowledged data-selection assumption that avoids the harder scale-ambiguity cases.
  • domain assumption The DUSt3R/Spann3R pretrained backbone provides generalizable pairwise 3D structure prediction, and the spatial memory maintains consistent canonical coordinates over long sequences.
    The whole method relies on the memory mechanism from Spann3R to align all views into the first view's coordinate frame without global optimization; the paper's own 50-view degradation suggests this assumption weakens with sequence length.
  • standard math The 3D Gaussian Splatting differentiable rasterizer correctly maps predicted Gaussians to images and is treated as a fixed, trustworthy module.
    The paper uses the standard 3D-GS formulation and rasterizer from prior work without re-deriving it; this is a reasonable background assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PreF3R: Pose-Free Feed-Forward 3D Gaussian Splatting from Variable-length Image Sequence." pith.science (2026). https://pith.science/paper/PVPLR5MG

@misc{pith2026241116877,
  author       = {Pith},
  title        = {Pith review of: PreF3R: Pose-Free Feed-Forward 3D Gaussian Splatting from Variable-length Image Sequence},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PVPLR5MG}},
  note         = {Machine review of arXiv:2411.16877}
}
read the original abstract

We present PreF3R, Pose-Free Feed-forward 3D Reconstruction from an image sequence of variable length. Unlike previous approaches, PreF3R removes the need for camera calibration and reconstructs the 3D Gaussian field within a canonical coordinate frame directly from a sequence of unposed images, enabling efficient novel-view rendering. We leverage DUSt3R's ability for pair-wise 3D structure reconstruction, and extend it to sequential multi-view input via a spatial memory network, eliminating the need for optimization-based global alignment. Additionally, PreF3R incorporates a dense Gaussian parameter prediction head, which enables subsequent novel-view synthesis with differentiable rasterization. This allows supervising our model with the combination of photometric loss and pointmap regression loss, enhancing both photorealism and structural accuracy. Given a sequence of ordered images, PreF3R incrementally reconstructs the 3D Gaussian field at 20 FPS, therefore enabling real-time novel-view rendering. Empirical experiments demonstrate that PreF3R is an effective solution for the challenging task of pose-free feed-forward novel-view synthesis, while also exhibiting robust generalization to unseen scenes.

Figures

Figures reproduced from arXiv: 2411.16877 by the authors.

Figure 1
Figure 1. Overview of PREF3R. Given a sequence of unposed images of variable length, PREF3R incrementally reconstructs a set of 3D Gaussian primitives in a single feed-forward pass without any pre-processing or intermediate pose estimation. PREF3R operates at 20 FPS on a single H100 GPU, enabling real-time novel-view synthesis from numerous input images through differentiable rasterization. Abstract We present PREF3R, Pose-Fr… view at source ↗
Figure 2
Figure 2. PREF3R’s overall architecture. Left: An ordered set of unposed images {It} T t=1 is fed into PREF3R sequentially. Middle: At timestamp t, the input frame It is first encoded by a ViT-encoder into ft, which is then decoded into the query feature f q t by the Target Decoder. The Target Decoder is intertwined with the Reference Decoder through cross-attention. Simultaneously, the query feature of the previous frame f q… view at source ↗
Figure 3
Figure 3. Scale ambiguity problem. Even slight scale shifts can cause significant view drifts in rendered results from ground-truth camera poses, making it hard to apply photometric supervision. Top row: ground-truth images; Bottom row: rendered images. Data sample is from Co3D [46]. 3.4. Training and inference Training. For pointmap regression, we use the same loss function as in DUSt3R [61], extending its formulation from p… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison of novel view synthesis performance. Left: visualization of scene reconstructions from ARK￾itScenes [6]; Right: visualization of reconstructions from ScanNet++ [74]. Each row corresponds to a unique viewpoint, while each column displays the outpu…
Figure 5
Figure 5. Figure 5: PREF3R performs incremental Gaussian reconstruction in real-time. Left: in-domain scene reconstruction from Scan￾Net++ [74]; Right: out-of-domain scene reconstruction from Tanks and Temples [36]. Method PSNR↑ SSIM↑ LPIPS↓ w/o extra views 22.27 0.788 0.131 w/o Gaussian …
Figure 6
Figure 6. Figure 6: Qualitative comparison of novel-view synthesis per￾formance. Top Row: When the input views are sufficiently dense, Spann3R produces high-quality projected images comparable to images rendered by PREF3R. Bottom Two Rows: In some cases, using colored pointmaps without Ga…
Figure 7
Figure 7. Figure 7: Qualitative comparison of cross-dataset novel-view synthesis performance. Left: visualization of 10-view scenes recon￾struction from ARKitScenes [6]; Ours‡ represents PREF3R trained on datasets excluding ARKitScenes. Right: visualization of 10-view scenes reconstructio…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 5 Pith papers

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

  1. TinySplat: Feedforward Approach for Generating Compact 3D Scene Representation

    cs.CV 2025-06 conditional novelty 7.0 of 10

    TinySplat compresses feedforward 3D Gaussian scenes by 105-199x on two-view benchmarks (about 50x on DL3DV) while keeping rendered quality close to the uncompressed model.

  2. PanoSplatt3R: Leveraging Perspective Pretraining for Generalized Unposed Wide-Baseline Panorama Reconstruction

    cs.CV 2025-07 conditional novelty 6.0 of 10

    PanoSplatt3R adapts a perspective pretrained stereo model to unposed wide-baseline panorama reconstruction with per-head rolled rotary positional embeddings, achieving SOTA on HM3D and Replica.

  3. SpatialTrackerV2: 3D Point Tracking Made Easy

    cs.CV 2025-07 conditional novelty 6.0 of 10

    A single feed-forward model jointly estimates video depth, camera poses, and 3D point trajectories from monocular video, setting a new state of the art on TAPVid-3D.

  4. E3D-Bench: A Benchmark for End-to-End 3D Geometric Foundation Models

    cs.CV 2025-06 conditional novelty 6.0 of 10

    E3D-Bench compares 16 3D geometric foundation models on depth, reconstruction, pose, and view-synthesis tasks with a unified evaluation toolkit.

  5. Review of Feed-forward 3D Reconstruction: From DUSt3R to VGGT

    cs.CV 2025-07 conditional novelty 3.0 of 10

    A survey of feed-forward 3D reconstruction models that jointly estimate camera poses and dense geometry from images in one network pass.

Reference graph

Works this paper leans on

80 extracted references · 56 canonical work pages · cited by 5 Pith papers

  1. [32]

    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 CVPR, pages 21357–21366, 2024. 6, 7, 1

    work page 2024

  2. [1]

    Building rome in a day

    Sameer Agarwal, Noah Snavely, Ian Simon, Steven M Seitz, and Richard Szeliski. Building rome in a day. InICCV, pages 72–79, 2009. 2

    work page 2009

  3. [2]

    Bundle adjustment in the large

    Sameer Agarwal, Noah Snavely, Steven M Seitz, and Richard Szeliski. Bundle adjustment in the large. In ECCV, pages 29–42, 2010. 2

    work page 2010

  4. [3]

    Barron, Ben Mildenhall, Matthew Tancik, Peter Hedman, Ricardo Martin-Brualla, and Pratul P

    Jonathan T. Barron, Ben Mildenhall, Matthew Tancik, Peter Hedman, Ricardo Martin-Brualla, and Pratul P. Srinivasan. Mip-nerf: A multiscale representation for anti-aliasing neu- ral radiance fields, 2021. 3

    work page 2021

  5. [4]

    Mip-nerf 360: Unbounded anti-aliased neural radiance fields

    Jonathan T Barron, Ben Mildenhall, Dor Verbin, Pratul P Srinivasan, and Peter Hedman. Mip-nerf 360: Unbounded anti-aliased neural radiance fields. In CVPR, pages 5470– 5479, 2022. 3

    work page 2022

  6. [5]

    Zip-nerf: Anti-aliased grid- based neural radiance fields

    Jonathan T Barron, Ben Mildenhall, Dor Verbin, Pratul P Srinivasan, and Peter Hedman. Zip-nerf: Anti-aliased grid- based neural radiance fields. In ICCV, pages 19697–19705,

  7. [6]

    Arkitscenes: A diverse real-world dataset for 3d indoor scene understanding using mobile rgb-d data

    Gilad Baruch, Zhuoyuan Chen, Afshin Dehghan, Yuri Fei- gin, Peter Fu, Thomas Gebauer, Daniel Kurz, Tal Dimry, Brandon Joffe, Arik Schwartz, et al. Arkitscenes: A diverse real-world dataset for 3d indoor scene understanding using mobile rgb-d data. In NeurIPS Datasets and Benchmarks ,

  8. [7]

    Codeslam—learning a compact, optimisable representation for dense visual slam

    Michael Bloesch, Jan Czarnowski, Ronald Clark, Stefan Leutenegger, and Andrew J Davison. Codeslam—learning a compact, optimisable representation for dense visual slam. In CVPR, pages 2560–2568, 2018. 2

    work page 2018

Show all 80 references
  1. [8]

    pixelsplat: 3d gaussian splats from image pairs for scalable generalizable 3d reconstruction, 2024

    David Charatan, Sizhe Li, Andrea Tagliasacchi, and Vincent Sitzmann. pixelsplat: 3d gaussian splats from image pairs for scalable generalizable 3d reconstruction, 2024. 3, 5, 6

    2024

  2. [9]

    Dbarf: Deep bundle-adjusting generalizable neural radiance fields, 2023

    Yu Chen and Gim Hee Lee. Dbarf: Deep bundle-adjusting generalizable neural radiance fields, 2023. 2, 3

    2023

  3. [10]

    Mvsplat: Efficient 3d gaussian splatting from sparse multi-view images

    Yuedong Chen, Haofei Xu, Chuanxia Zheng, Bohan Zhuang, Marc Pollefeys, Andreas Geiger, Tat-Jen Cham, and Jianfei Cai. Mvsplat: Efficient 3d gaussian splatting from sparse multi-view images. arXiv preprint arXiv:2403.14627, 2024. 3, 5, 6, 7, 1, 2

    2024 arXiv

  4. [11]

    Garf: Gaussian activated radiance fields for high fidelity reconstruction and pose estimation, 2022

    Shin-Fang Chng, Sameera Ramasinghe, Jamie Sherrah, and Simon Lucey. Garf: Gaussian activated radiance fields for high fidelity reconstruction and pose estimation, 2022. 3

    2022

  5. [12]

    Discrete-continuous optimization for large- scale structure from motion

    David Crandall, Andrew Owens, Noah Snavely, and Dan Huttenlocher. Discrete-continuous optimization for large- scale structure from motion. In CVPR, pages 3001–3008,

  6. [13]

    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 CVPR, pages 5828–5839, 2017. 6, 1

    2017

  7. [14]

    Monoslam: Real-time single camera slam

    Andrew J Davison, Ian D Reid, Nicholas D Molton, and Olivier Stasse. Monoslam: Real-time single camera slam. TPAMI, 29(6):1052–1067, 2007. 2

    2007

  8. [15]

    Superpoint: Self-supervised interest point detection and description

    Daniel DeTone, Tomasz Malisiewicz, and Andrew Rabi- novich. Superpoint: Self-supervised interest point detection and description. In CVPRW, pages 224–236, 2018. 2

    2018

  9. [16]

    Learning a depth covariance function

    Eric Dexheimer and Andrew J Davison. Learning a depth covariance function. In CVPR, pages 13122–13131, 2023. 3

    2023

  10. [17]

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

    Carl Doersch, Yi Yang, Mel Vecerik, Dilara Gokay, Ankush Gupta, Yusuf Aytar, Joao Carreira, and Andrew Zisserman. Tapir: Tracking any point with per-frame initialization and temporal refinement. In ICCV, pages 10061–10072, 2023. 3

    2023

  11. [18]

    An image is worth 16x16 words: Trans- formers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, et al. An image is worth 16x16 words: Trans- formers for image recognition at scale. In ICLR, 2020. 2, 7

    2020

  12. [19]

    Deep- videomvs: Multi-view stereo on video with recurrent spatio- temporal fusion

    Arda Duzceker, Silvano Galliani, Christoph V ogel, Pablo Speciale, Mihai Dusmanu, and Marc Pollefeys. Deep- videomvs: Multi-view stereo on video with recurrent spatio- temporal fusion. In CVPR, pages 15324–15333, 2021. 3

    2021

  13. [20]

    Instantsplat: Sparse-view sfm-free gaussian splatting in seconds, 2024

    Zhiwen Fan, Wenyan Cong, Kairun Wen, Kevin Wang, Jian Zhang, Xinghao Ding, Danfei Xu, Boris Ivanovic, Marco Pavone, Georgios Pavlakos, Zhangyang Wang, and Yue Wang. Instantsplat: Sparse-view sfm-free gaussian splatting in seconds, 2024. 3, 7, 1

    2024

  14. [21]

    Large spatial model: End-to-end unposed images to semantic 3d, 2024

    Zhiwen Fan, Jian Zhang, Wenyan Cong, Peihao Wang, Renjie Li, Kairun Wen, Shijie Zhou, Achuta Kadambi, Zhangyang Wang, Danfei Xu, Boris Ivanovic, Marco Pavone, and Yue Wang. Large spatial model: End-to-end unposed images to semantic 3d, 2024. 2, 3

    2024

  15. [22]

    Accurate, dense, and ro- bust multiview stereopsis

    Yasutaka Furukawa and Jean Ponce. Accurate, dense, and ro- bust multiview stereopsis. TPAMI, 32(8):1362–1376, 2009. 2

    2009

  16. [23]

    Massively parallel multiview stereopsis by surface normal diffusion

    Silvano Galliani, Katrin Lasinger, and Konrad Schindler. Massively parallel multiview stereopsis by surface normal diffusion. In ICCV, pages 873–881, 2015. 2

    2015

  17. [24]

    Detector-free struc- ture from motion

    Xingyi He, Jiaming Sun, Yifan Wang, Sida Peng, Qixing Huang, Hujun Bao, and Xiaowei Zhou. Detector-free struc- ture from motion. CVPR, 2024. 2

    2024

  18. [25]

    Unifying corre- 9 spondence, pose and nerf for pose-free novel view synthesis from stereo pairs

    Sunghwan Hong, Jaewoo Jung, Heeseong Shin, Jiaolong Yang, Seungryong Kim, and Chong Luo. Unifying corre- 9 spondence, pose and nerf for pose-free novel view synthesis from stereo pairs. arXiv preprint arXiv:2312.07246, 2023. 2

    2023 arXiv

  19. [26]

    2d gaussian splatting for geometrically accu- rate radiance fields

    Binbin Huang, Zehao Yu, Anpei Chen, Andreas Geiger, and Shenghua Gao. 2d gaussian splatting for geometrically accu- rate radiance fields. In ACM SIGGRAPH, pages 1–11, 2024. 3

    2024

  20. [27]

    Robusti- fying the multi-scale representation of neural radiance fields,

    Nishant Jain, Suryansh Kumar, and Luc Van Gool. Robusti- fying the multi-scale representation of neural radiance fields,

  21. [28]

    Codenerf: Disentan- gled neural radiance fields for object categories

    Wonbong Jang and Lourdes Agapito. Codenerf: Disentan- gled neural radiance fields for object categories. In ICCV, pages 12949–12958, 2021. 3

    2021

  22. [29]

    Nvist: In the wild new view synthesis from a single image with transformers

    Wonbong Jang and Lourdes Agapito. Nvist: In the wild new view synthesis from a single image with transformers. In CVPR, pages 10181–10193, 2024. 3

    2024

  23. [30]

    Co- tracker: It is better to track together

    Nikita Karaev, Ignacio Rocco, Benjamin Graham, Natalia Neverova, Andrea Vedaldi, and Christian Rupprecht. Co- tracker: It is better to track together. arXiv preprint arXiv:2307.07635, 2023. 3

    2023 arXiv

  24. [31]

    Repurpos- ing diffusion-based image generators for monocular depth estimation

    Bingxin Ke, Anton Obukhov, Shengyu Huang, Nando Met- zger, Rodrigo Caye Daudt, and Konrad Schindler. Repurpos- ing diffusion-based image generators for monocular depth estimation. In CVPR, pages 9492–9502, 2024. 3

    2024

  25. [33]

    3d gaussian splatting for real-time radiance field rendering

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk ¨uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering. TOG, 42(4):139–1, 2023. 3, 4, 5, 6

    2023

  26. [34]

    3d gaussian splatting for real-time radiance field rendering, 2023

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk ¨uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering, 2023. 2

    2023

  27. [35]

    Parallel tracking and map- ping for small ar workspaces

    Georg Klein and David Murray. Parallel tracking and map- ping for small ar workspaces. In ISMAR, pages 1–10, 2007. 2

    2007

  28. [36]

    Tanks and temples: Benchmarking large-scale scene reconstruction

    Arno Knapitsch, Jaesik Park, Qian-Yi Zhou, and Vladlen Koltun. Tanks and temples: Benchmarking large-scale scene reconstruction. ACM Transactions on Graphics, 36(4), 2017. 8

    2017

  29. [37]

    Ground- ing image matching in 3d with mast3r, 2024

    Vincent Leroy, Yohann Cabon, and J´erˆome Revaud. Ground- ing image matching in 3d with mast3r, 2024. 2, 3, 4, 6

    2024

  30. [38]

    Barf: Bundle-adjusting neural radiance fields,

    Chen-Hsuan Lin, Wei-Chiu Ma, Antonio Torralba, and Si- mon Lucey. Barf: Bundle-adjusting neural radiance fields,

  31. [39]

    Infinite na- ture: Perpetual view generation of natural scenes from a sin- gle image

    Andrew Liu, Richard Tucker, Varun Jampani, Ameesh Makadia, Noah Snavely, and Angjoo Kanazawa. Infinite na- ture: Perpetual view generation of natural scenes from a sin- gle image. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2021. 1

    2021

  32. [40]

    Gnerf: Gan-based neu- ral radiance field without posed camera, 2021

    Quan Meng, Anpei Chen, Haimin Luo, Minye Wu, Hao Su, Lan Xu, Xuming He, and Jingyi Yu. Gnerf: Gan-based neu- ral radiance field without posed camera, 2021. 3

    2021

  33. [41]

    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 ECCV, pages 405–421, 2020. 2, 3

    2020

  34. [42]

    Instant neural graphics primitives with a multires- olution hash encoding

    Thomas M ¨uller, Alex Evans, Christoph Schied, and Alexan- der Keller. Instant neural graphics primitives with a multires- olution hash encoding. ACM Trans. Graph. , 41(4):102:1– 102:15, 2022. 2

    2022

  35. [43]

    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 ICCV, pages 2320–2327, 2011. 2

    2011

  36. [44]

    Sch¨onberger

    Linfei Pan, D ´aniel Bar´ath, Marc Pollefeys, and Johannes L. Sch¨onberger. Global structure-from-motion revisited, 2024. 2, 3

    2024

  37. [45]

    Vi- sion transformers for dense prediction

    Ren ´e Ranftl, Alexey Bochkovskiy, and Vladlen Koltun. Vi- sion transformers for dense prediction. In ICCV, pages 12179–12188, 2021. 2, 5, 7

    2021

  38. [46]

    Com- mon objects in 3d: Large-scale learning and evaluation of real-life 3d category reconstruction

    Jeremy Reizenstein, Roman Shapovalov, Philipp Henzler, Luca Sbordone, Patrick Labatut, and David Novotny. Com- mon objects in 3d: Large-scale learning and evaluation of real-life 3d category reconstruction. In ICCV, pages 10901– 10911, 2021. 5

    2021

  39. [47]

    Superglue: Learning feature matching with graph neural networks

    Paul-Edouard Sarlin, Daniel DeTone, Tomasz Malisiewicz, and Andrew Rabinovich. Superglue: Learning feature matching with graph neural networks. InCVPR, pages 4938– 4947, 2020. 2

    2020

  40. [48]

    Simplere- con: 3d reconstruction without 3d convolutions

    Mohamed Sayed, John Gibson, Jamie Watson, Victor Prisacariu, Michael Firman, and Cl´ement Godard. Simplere- con: 3d reconstruction without 3d convolutions. In ECCV, pages 1–19, 2022. 3

    2022

  41. [49]

    Structure- from-motion revisited

    Johannes L Schonberger and Jan-Michael Frahm. Structure- from-motion revisited. In CVPR, pages 4104–4113, 2016. 2

    2016

  42. [50]

    Pixelwise view selection for unstructured multi-view stereo

    Johannes L Sch ¨onberger, Enliang Zheng, Jan-Michael Frahm, and Marc Pollefeys. Pixelwise view selection for unstructured multi-view stereo. In ECCV, pages 501–518,

  43. [51]

    Splatt3r: Zero-shot gaussian splatting from uncalibrated image pairs, 2024

    Brandon Smart, Chuanxia Zheng, Iro Laina, and Vic- tor Adrian Prisacariu. Splatt3r: Zero-shot gaussian splatting from uncalibrated image pairs, 2024. 2, 3

    2024

  44. [52]

    Photo tourism: exploring photo collections in 3d

    Noah Snavely, Steven M Seitz, and Richard Szeliski. Photo tourism: exploring photo collections in 3d. TOG, 25(3):835– 846, 2006. 2

    2006

  45. [53]

    Loftr: Detector-free local feature matching with transformers

    Jiaming Sun, Zehong Shen, Yuang Wang, Hujun Bao, and Xiaowei Zhou. Loftr: Detector-free local feature matching with transformers. In CVPR, pages 8922–8931, 2021. 2

    2021

  46. [54]

    Optimizing the viewing graph for structure-from-motion

    Chris Sweeney, Torsten Sattler, Tobias Hollerer, Matthew Turk, and Marc Pollefeys. Optimizing the viewing graph for structure-from-motion. In ICCV, pages 801–809, 2015. 2

    2015

  47. [55]

    Splatter image: Ultra-fast single-view 3d recon- struction, 2024

    Stanislaw Szymanowicz, Christian Rupprecht, and Andrea Vedaldi. Splatter image: Ultra-fast single-view 3d recon- struction, 2024. 3

    2024

  48. [56]

    Raft: Recurrent all-pairs field transforms for optical flow

    Zachary Teed and Jia Deng. Raft: Recurrent all-pairs field transforms for optical flow. In ECCV, pages 402–419, 2020. 3

    2020

  49. [57]

    Bundle adjustment—a modern synthe- sis

    Bill Triggs, Philip F McLauchlan, Richard I Hartley, and An- drew W Fitzgibbon. Bundle adjustment—a modern synthe- sis. In ICCVW, pages 298–372, 2000. 2

    2000

  50. [58]

    3d reconstruction with spatial memory, 2024

    Hengyi Wang and Lourdes Agapito. 3d reconstruction with spatial memory, 2024. 2, 3, 4, 6, 7, 1 10

    2024

  51. [59]

    Vggsfm: Visual geometry grounded deep structure from motion

    Jianyuan Wang, Nikita Karaev, Christian Rupprecht, and David Novotny. Vggsfm: Visual geometry grounded deep structure from motion. In CVPR, pages 21686–21697, 2024. 2

    2024

  52. [60]

    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. In NIPS, pages 27171–27183, 2021. 3

    2021

  53. [61]

    Dust3r: Geometric 3d vi- sion made easy

    Shuzhe Wang, Vincent Leroy, Yohann Cabon, Boris Chidlovskii, and Jerome Revaud. Dust3r: Geometric 3d vi- sion made easy. In CVPR, pages 20697–20709, 2024. 2, 3, 4, 5, 6

    2024

  54. [62]

    Nerf–: Neural radiance fields without known camera parameters, 2022

    Zirui Wang, Shangzhe Wu, Weidi Xie, Min Chen, and Vic- tor Adrian Prisacariu. Nerf–: Neural radiance fields without known camera parameters, 2022. 2, 3

    2022

  55. [63]

    Robust global translations with 1dsfm

    Kyle Wilson and Noah Snavely. Robust global translations with 1dsfm. In ECCV, pages 61–75, 2014. 2

    2014

  56. [64]

    Towards linear-time incremental structure from motion

    Changchang Wu. Towards linear-time incremental structure from motion. In 3DV, pages 127–134, 2013. 2

    2013

  57. [65]

    Multicore bundle adjustment

    Changchang Wu, Sameer Agarwal, Brian Curless, and Steven M Seitz. Multicore bundle adjustment. In CVPR, pages 3057–3064, 2011. 2

    2011

  58. [66]

    Sinerf: Sinusoidal neural radiance fields for joint pose esti- mation and scene reconstruction, 2022

    Yitong Xia, Hao Tang, Radu Timofte, and Luc Van Gool. Sinerf: Sinusoidal neural radiance fields for joint pose esti- mation and scene reconstruction, 2022. 3

    2022

  59. [67]

    Spatialtracker: Tracking any 2d pixels in 3d space

    Yuxi Xiao, Qianqian Wang, Shangzhan Zhang, Nan Xue, Sida Peng, Yujun Shen, and Xiaowei Zhou. Spatialtracker: Tracking any 2d pixels in 3d space. In CVPR, pages 20406– 20417, 2024. 3

    2024

  60. [68]

    Sparp: Fast 3d object reconstruction and pose estimation from sparse views, 2024

    Chao Xu, Ang Li, Linghao Chen, Yulin Liu, Ruoxi Shi, Hao Su, and Minghua Liu. Sparp: Fast 3d object reconstruction and pose estimation from sparse views, 2024. 2

    2024

  61. [69]

    Carff: Conditional auto-encoded radiance field for 3d scene forecasting, 2024

    Jiezhi Yang, Khushi Desai, Charles Packer, Harshil Bhatia, Nicholas Rhinehart, Rowan McAllister, and Joseph Gonza- lez. Carff: Conditional auto-encoded radiance field for 3d scene forecasting, 2024. 3

    2024

  62. [70]

    Mvsnet: Depth inference for unstructured multi-view stereo

    Yao Yao, Zixin Luo, Shiwei Li, Tian Fang, and Long Quan. Mvsnet: Depth inference for unstructured multi-view stereo. In ECCV, pages 767–783, 2018. 2, 3

    2018

  63. [71]

    V ol- ume rendering of neural implicit surfaces

    Lior Yariv, Jiatao Gu, Yoni Kasten, and Yaron Lipman. V ol- ume rendering of neural implicit surfaces. In NIPS, pages 4805–4815, 2021. 3

    2021

  64. [72]

    No pose, no problem: Surprisingly simple 3d gaussian splats from sparse unposed images, 2024

    Botao Ye, Sifei Liu, Haofei Xu, Xueting Li, Marc Pollefeys, Ming-Hsuan Yang, and Songyou Peng. No pose, no problem: Surprisingly simple 3d gaussian splats from sparse unposed images, 2024. 2, 3

    2024

  65. [73]

    gsplat: An open-source library for Gaussian splatting

    Vickie Ye, Ruilong Li, Justin Kerr, Matias Turkulainen, Brent Yi, Zhuoyang Pan, Otto Seiskari, Jianbo Ye, Jeffrey Hu, Matthew Tancik, and Angjoo Kanazawa. gsplat: An open-source library for Gaussian splatting. arXiv preprint arXiv:2409.06765, 2024. 6

    2024 arXiv

  66. [74]

    Scannet++: A high-fidelity dataset of 3d indoor scenes

    Chandan Yeshwanth, Yueh-Cheng Liu, Matthias Nießner, and Angela Dai. Scannet++: A high-fidelity dataset of 3d indoor scenes. In ICCV, 2023. 6, 7, 8, 1, 2, 3

    2023

  67. [75]

    Metric3d: Towards zero-shot metric 3d prediction from a single image

    Wei Yin, Chi Zhang, Hao Chen, Zhipeng Cai, Gang Yu, Kaixuan Wang, Xiaozhi Chen, and Chunhua Shen. Metric3d: Towards zero-shot metric 3d prediction from a single image. In ICCV, pages 9043–9053, 2023. 2, 3

    2023

  68. [76]

    Sim-sync: From certifiably op- timal synchronization over the 3d similarity group to scene reconstruction with learned depth

    Xihang Yu and Heng Yang. Sim-sync: From certifiably op- timal synchronization over the 3d similarity group to scene reconstruction with learned depth. IEEE Robotics and Au- tomation Letters, 2024. 2

    2024

  69. [77]

    Stereo magnification: Learning view synthesis using multiplane images

    Tinghui Zhou, Richard Tucker, John Flynn, Graham Fyffe, and Noah Snavely. Stereo magnification: Learning view synthesis using multiplane images. arXiv preprint arXiv:1805.09817, 2018. 1 11 Supplementary Material Spann3R Ours GT Figure 6. Qualitative comparison of novel-view sy...

    2018 arXiv

  70. [78]

    4.1, we construct several baselines from existing methods

    More implementation details As described in Sec. 4.1, we construct several baselines from existing methods. Splatt3R [32] is a pose-free feed- forward Gaussian model taking pairwise images as input. The original model is trained on both training and validation splits of ScanNe...

  71. [79]

    More experimental results We provide a qualitative comparison with Spann3R [58] in Fig. 6. Although we share a similar structural model with Spann3R, the rendering results of P REF3R surpasses Spann3R significantly. This is because using colored pointmaps without Gaussian para...

  72. [80]

    First, PREF3R is trained on image frames with dense input views, with substantial overlap between adjacent frames

    Limitations and future works Through empirical experiments, P REF3R demonstrates competitive efficiency and rendering quality across a variety of datasets, but it retains several limitations. First, PREF3R is trained on image frames with dense input views, with substantial ove...

Pith tools

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