Pith. sign in

REVIEW 3 major objections 4 minor 1 cited by

Elastic3D claims flat video becomes controllable stereo with no depth or warping: one-step latent diffusion generates the right-eye view, steered by a disparity knob and a left-view-guided decoder, setting state-of-the-art numbers on three

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-03 16:12 UTC pith:XMJE6KRC

load-bearing objection Solid engineering with an honest limitations section, but the 'new state of the art' claim only holds under an oracle-conditioning evaluation; cite the guided decoder and disparity-control mechanism, not the SOTA tables as-is. the 3 major comments →

arxiv 2512.14236 v2 pith:XMJE6KRC submitted 2025-12-16 cs.CV

Elastic3D: Controllable Stereo Video Conversion with Guided Latent Decoding

Nando Metzger , Prune Truong , Goutam Bhat , Konrad Schindler , Federico Tombari This is my paper
classification cs.CV
keywords stereo video conversionmonocular-to-stereolatent diffusiondisparity conditioningguided VAE decoderepipolar attentionbinocular rivalrywarping-free synthesis
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper tries to establish that stereo video conversion does not need a depth-estimation-and-warping stage at all. Instead, a one-step latent diffusion model, Elastic3D, generates the right-eye video directly from the left-eye video, conditioned on a single scalar that sets the median disparity — an intuitive '3D strength' knob the user can turn at inference. To overcome the blurriness and texture loss of latent video diffusion, the authors add a guided VAE decoder that pulls high-frequency detail from the left view along epipolar lines, which they show sharply reduces binocular rivalry. On three in-the-wild stereo video datasets, the method reports the best numbers across a new four-part evaluation protocol (overall quality, stereoscopic fidelity, geometric correctness, temporal stability), and a headset user study prefers it over two published baselines. If correct, it means ordinary flat video could be turned into comfortable, controllable 3D content in a single fast forward pass — with the caveat that the evaluation hands the model its target median disparity.

Core claim

The paper's central claim is that the right-eye view of a stereo video can be synthesized directly from the left-eye video, without explicit depth estimation or geometric warping. Elastic3D does this with a one-step latent diffusion model built on a pretrained video-diffusion backbone: it takes the compressed left-view latent and a scalar 'median disparity' token, and generates the right-view latent in a single forward pass. Two components carry the argument: the scalar token gives continuous, intuitive control over the output disparity range (so one model covers different camera baselines), and a guided VAE decoder re-injects high-frequency detail from the left video through epipolar-constr

What carries the argument

The load-bearing mechanism is the guided latent-decoding loop. A frozen VAE encoder compresses the left video into latents; a 1-step denoising U-Net — trained as a feed-forward generator with zero noise at the maximum timestep — outputs the right-view latent; and a guided decoder, initialized from the standard VAE decoder, reconstructs the right view while a guidance network (initialized from the VAE encoder) supplies multi-scale features of the left video. Inside each decoder upsampling block, a zero-initialized epipolar cross-attention layer lets each decoder feature attend only to the corresponding horizontal scanline (the epipolar line in the rectified case) of the left-view feature map,

Load-bearing premise

At evaluation time the model is told the correct median disparity of the target pair — the paper states that 'the GT disparity is used to compute the conditioning factor δ' — so the headline results assume the desired 3D strength is known rather than discovered from the monocular input; if the user or an automatic estimator sets δ wrongly, the alignment and quality gains may shrink.

What would settle it

Take the same left-eye videos and feed them through Elastic3D with the conditioning knob set by an automatic estimate (for example, the median of a monocular depth map rescaled into pixels) instead of the ground-truth median disparity, then recompute the disparity error using an independent stereo matcher different from the one that generated the pseudo-ground-truth. If PSNR drops by several dB or disparity errors grow markedly under that honest, fully monocular protocol, then the reported state of the art comes from oracle conditioning and evaluator self-consistency, not from strength of the

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Monocular video libraries could be converted to stereo with a user-adjustable 3D intensity, allowing comfortable viewing on headsets with different interocular settings.
  • Removing the explicit depth-and-warp stage eliminates disocclusion holes and depth-estimator failure modes, so thin structures, text, and non-Lambertian surfaces survive in the generated view.
  • Sharp texture is transferred from the source eye to the target eye, cutting the mismatch that causes binocular rivalry and making long stereo viewing more comfortable.
  • One-step feed-forward inference (about 1.7 seconds for 16 512x512 frames on an H100 GPU) makes near-interactive conversion practical, whereas multi-stage depth-plus-diffusion pipelines take roughly 5 seconds or more.
  • The four-part evaluation protocol gives future stereo-conversion systems a common yardstick covering quality, fidelity, geometry, and temporal stability.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Fully automatic conversion will require estimating δ from the left view or from user intent; the paper does not evaluate that setting, so the 'monocular' framing should be read as 'monocular plus a target-disparity hint.'
  • The global scalar is a coarse control: the paper's own limitation section observes that a background-dominated scene can have median disparity near zero while foreground objects demand nonzero disparity. A per-frame or spatially varying conditioning signal is a natural testable extension for zoom and dolly shots.
  • Because the guided decoder is trained separately and plugs into other latent diffusion pipelines, its benefit may transfer beyond stereo conversion, e.g., to any two-view synthesis task where one view must borrow detail from another along known epipolar geometry.
  • The 1-D epipolar attention presumes rectified stereo; applying the approach to unrectified, rotating, or rolling cameras would require generalizing the attention search to full epipolar lines in arbitrary directions.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper presents Elastic3D, a warping-free latent diffusion model for monocular-to-stereo video conversion. The method synthesizes the right-eye video directly from the left video using a one-step denoising process, conditioned on a scalar median-disparity token that provides user-controllable 3D strength. A guided VAE decoder with epipolar attention re-injects high-frequency detail from the left view to reduce binocular rivalry. The authors evaluate on Stereo4D, Ego4D, and the Spatial Video dataset (AVP and iPhone), reporting favorable metrics against warp-based and warping-free baselines, plus a small headset user study. The central claim is that Elastic3D sets a new state of the art for controllable stereo video conversion.

Significance. If the results hold, the paper makes a solid contribution: a feed-forward, single-step warping-free stereo video generator with intuitive scalar disparity control, a plug-and-play guided decoder that improves detail preservation, and a more complete evaluation protocol covering geometry, stereoscopic fidelity, and temporal consistency. The ablation study is thorough, the user study is a genuinely useful qualitative check, and the supplementary material is unusually transparent about limitations, including the ambiguity of median conditioning and occasional geometric hallucinations. The main caveat is that the quantitative SOTA claim is evaluated under an oracle condition: Elastic3D receives the ground-truth median disparity at inference, and the geometric metric relies on the same stereo estimator family used to produce training labels. This does not make the method internally inconsistent, but it narrows what Tables 5–7 certify. The claim of a 'new standard' should be re-scoped to 'given a target median disparity' unless automatic disparity estimation is also evaluated.

major comments (3)
  1. [Sec. 6.1 / Tables 5–7] The main SOTA comparison gives Elastic3D the ground-truth median disparity δ at inference ('For our method, the GT disparity is used to compute the conditioning factor δ'), while the abstract and conclusion present the method as automated monocular-to-stereo conversion. Since PSNR/SSIM/LPIPS strongly penalize global disparity misalignment, providing the exact target δ directly controls the dominant error term. Baselines receive scale/shift-aligned monocular depth, which is also target-derived, but the asymmetry remains important. Please either (a) evaluate with an estimated or user-chosen δ and report sensitivity to δ errors, or (b) re-scope the headline claim as 'given a user-specified median disparity.' The paper's own Sec. H acknowledges ambiguity in median conditioning; that limitation should be reflected in the abstract/conclusion.
  2. [Sec. 5.1 / Sec. 6.1] The Disp.err metric estimates dense disparity on generated views with FoundationStereo and compares against pseudo-GT disparities also produced by FoundationStereo. The same estimator family supplies the training labels for δ and the synthetic warping augmentation. This self-consistency can reward outputs that mimic FoundationStereo's error distribution rather than true geometric accuracy. The paper should validate the geometric claim with an independent stereo matcher (e.g., RAFT-Stereo or a learned matcher not involved in training) on at least a subset, or explicitly discuss this limitation. The current text calls the reference 'pseudo' but does not analyze the circularity.
  3. [Tab. 1 (iPhone row) / Tab. 13] The text states that the proposed conditioning 'does not impact the geometric accuracy (Disp. err)', but the iPhone row of Table 1 shows Disp.err worsening from 0.64 without conditioning to 0.77 with conditioning — a roughly 20% relative degradation. The supplementary Table 13 repeats the same numbers. If this is noise, variance, or a side effect of the different training distribution, the claim should be qualified; as written, the table contradicts the stated conclusion.
minor comments (4)
  1. [Tab. 4] In the guided-decoder ablation, M2SVid + D′ achieves a lower Matchability Error (24.8) than Elastic3D (27.8). The text emphasizes a 'drastic reduction' for Elastic3D but does not mention that the plug-and-play decoder on a warp-based baseline outperforms the full warping-free model on this particular metric. This should be acknowledged for fairness.
  2. [Sec. 5 / evaluation protocol] The protocol is called 'black-box' but provides target-derived global 3D information to every method. This is a reasonable controlled-strength design, but the terminology is misleading and should be replaced with something like 'controlled-strength pipeline evaluation'.
  3. [Sec. E.1 / Table 9] The user study uses 9 participants and 120 pairwise judgments per comparison. The 50.0% vs 12.5% preference over Eye2Eye is plausible, but no significance testing or inter-participant variability is reported. Adding a confidence interval or a simple sign test would strengthen the claim.
  4. [References] Reference [22] is marked 'Submitted' and the Spatial Video Dataset is central to the evaluation. Please update the reference or indicate the arXiv/version status. Also, there are minor typos such as 'simplidied' (Supp. Sec. C) and 'Cputer Vision' in reference [66].

Circularity Check

0 steps flagged

No construction-level circularity; SOTA claim is scope-limited by oracle median-disparity conditioning and same-estimator pseudo-GT geometry evaluation.

full rationale

Elastic3D's construction-level derivation is self-contained and non-circular. The synthesis core is a 1-step latent diffusion map f_theta(0, z_L, tau(delta)) trained with latent and pixel-space losses (Eqs. 1, 2, 11, 12); the generated right view is not an algebraic consequence of the conditioning scalar delta or the left video alone. The guided decoder is trained as a reconstruction module, ablated both in isolation (Tab. 3) and on the full task (Tab. 4), and is initialized to preserve the pretrained SVD decoder mapping via zero-initialized output projections, so its gains are empirical rather than assumed. The M2SVid [50] citation is used for data splits, data handling, and as a baseline; it shares three authors with the present paper, but it is not invoked as a uniqueness theorem or as the justification for the conditioning or decoder design, so the self-citation is not load-bearing. Two evaluation-scope issues prevent the headline SOTA claim from being fully certified from monocular input alone, but neither reduces by construction. Sec. 5 states that all methods 'operate solely on the monocular input I_L at inference time', yet Sec. 6.1 says 'For our method, the GT disparity is used to compute the conditioning factor delta with eq. (3)': the target median disparity is supplied as an oracle or user-set knob rather than predicted. Second, the Disp.err metric estimates disparity on generated views with FoundationStereo and compares against pseudo-GT disparity also produced by FoundationStereo (Sec. 5.1 vs 6.1), while the same estimator family supplied the training labels for delta; this self-consistency can reward outputs that mimic FoundationStereo's error distribution. Sec. H openly lists the ambiguity of median conditioning and geometric hallucinations, confirming the authors' awareness of these scope limits. These are correctness/scope concerns, not construction-level circularity.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 0 invented entities

No new particles, forces, dimensions, or physical entities. The 'guided decoder' and 'disparity token' are engineered modules composed of existing operations; they have no independent falsifiable handle outside the system. The central claim rests on standard ML assumptions plus hand-chosen hyperparameters and an oracle conditioning signal.

free parameters (5)
  • test-time median disparity δ (oracle conditioning) = GT/FoundationStereo median disparity of target pair in all quantitative comparisons
    At inference δ is user-provided; in the paper's experiments it is computed from the target stereo pair's disparity (Eq. 3 + Sec 6.1), so the '3D strength' is given rather than estimated from the left video. Central to claimed controllability and pixel alignment.
  • disparity scaling factor for conditioning = ≈10^-2
    Found empirically 'crucial for numerical stability' (App. A.2); affects conditioning token scale.
  • synthetic augmentation disparity scale set = {0.05,0.1,0.2,0.4,0.6,0.8,1.25,1.5,2.0,3.0}
    Hand-chosen discrete scales used to create synthetic right views by forward warping; widens conditioning range (App. A.1). Central to teaching out-of-baseline disparity control.
  • zero-disparity augmentation probability = 1%
    Hand-chosen; teaches the δ=0 mapping.
  • loss weights = equal weights (L2 latent; L1, SSIM, LPIPS pixel; decoder L1+LPIPS)
    Chosen by hand; no sensitivity analysis reported for loss weighting.
axioms (5)
  • domain assumption Rectified stereo geometry: epipolar lines are horizontal, so stereo correspondence is a 1D search along rows.
    Used in guided decoder epipolar attention (Sec 4.3, Eq. 6-7) and in the P-PSNR/metric design. If the input/target rig is not rectified, the mechanism breaks.
  • domain assumption FoundationStereo's disparity estimates are a valid pseudo-ground truth for training labels, conditioning, synthetic warping, and the Disp.err metric.
    Sec 6.1/5.1; no independent geometric validation. The geometric-correctness claim is therefore a consistency check against a learned stereo matcher, not against measured depth.
  • standard math 1-step zero-noise diffusion (z_T≈0, v≈−z) is a valid feed-forward approximation of the pretrained SVD denoiser.
    Sec. C uses [12] to justify training fθ(0,T,z_L,τ(δ)) as a direct predictor of −v; depends on αbar_T≈0 and zero noise expectation.
  • ad hoc to paper A single scalar median disparity is a sufficient conditioning signal to control global 3D strength.
    Eq. (3) and Sec 4.2. The authors themselves list the ambiguity as a limitation (App. H): background-dominated scenes with δ≈0 give weak/incorrect conditioning when foreground disparity changes.
  • ad hoc to paper Synthetic forward-warped pairs (left warped by scaled pseudo-GT disparity) are a valid proxy for real stereo pairs of different baselines.
    App. A.1; used to teach out-of-distribution disparity ranges. Assumes the estimated disparity maps and simple forward warping produce plausible target views; invalid pixels are masked.

pith-pipeline@v1.3.0-alltime-deepseek · 38572 in / 15774 out tokens · 131402 ms · 2026-08-03T16:12:36.627857+00:00 · methodology

0 comments
read the original abstract

The growing demand for immersive 3D content calls for automated monocular-to-stereo video conversion. We present Elastic3D, a controllable, direct end-to-end method for upgrading a conventional video to a binocular one. Our approach, based on (conditional) latent diffusion, avoids artifacts due to explicit depth estimation and warping. The key to its high-quality stereo video output is a novel, guided VAE decoder that ensures sharp and epipolar-consistent stereo video output. Moreover, our method gives the user control over the strength of the stereo effect (more precisely, the disparity range) at inference time, via an intuitive, scalar tuning knob. Experiments on three different datasets of real-world stereo videos show that our method outperforms both traditional warping-based and recent warping-free baselines and sets a new standard for reliable, controllable stereo video conversion. Please check the project page for the video samples https://elastic3d.github.io.

Figures

Figures reproduced from arXiv: 2512.14236 by Federico Tombari, Goutam Bhat, Konrad Schindler, Nando Metzger, Prune Truong.

Figure 1
Figure 1. Figure 1: We introduce a direct, warping-free and feed-forward approach for mono-to-stereo video conversion. Our latent diffusion model [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Inference Pipeline. A frozen VAE Encoder E computes the latent code zL of the input video VL. The synthesis network fθ then generates the right-view latent zˆR, conditioned on zL and on a 3D strength control token τ (δ) (Sec. 4.2). Finally, our Guided Decoder D ′ (Sec. 4.3) renders the high-fidelity output IˆR, using both the generated latent zˆR and the original video VL as guidance. inherently provide st… view at source ↗
Figure 3
Figure 3. Figure 3: The strength of the stereo effect can be controlled by [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Decoding the ground truth latent. The compression with [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparison. Our method generates better textures with high-frequency details. The geometry of the generated stereo sample is also more correct – in line with the ground truth disparity. Note: This figure has been compressed for arXiv submission. Please refer to the project page for the uncompressed version. Results on Stereo4D: ReStereo, which combines video generation with restoration, often h… view at source ↗
Figure 6
Figure 6. Figure 6: Robustness to Misalignment. Sensitivity of Standard PSNR vs. our PatchPSNR (P-PSNR) to horizontal pixel shifts. P￾PSNR remains robust to small geometric errors, focusing on tex￾ture quality [PITH_FULL_IMAGE:figures/full_fig_p015_6.png] view at source ↗
Figure 9
Figure 9. Figure 9: Sensitivity to Blur. Sensitivity of SSIM vs. our Matcha￾bility Error to Gaussian blur. Matchability is more sensitive to blur than SSIM, effectively penalizing over-smoothed generations. icantly more sensitive to the loss of high-frequency details than SSIM. A blurry image might still look “structurally” similar (high SSIM), but it will fail to produce reliable key￾points, leading to a high Matchability Er… view at source ↗
Figure 10
Figure 10. Figure 10: Visualization of Matchability Error Components. Green: TP (Intersection), Orange: FP (Hallucination), Red: FN (Omis￾sion). Visualizing the error components reveals that M2SVid yields a higher error driven by significant omission of details (Red). We note that our model produces slightly more hallucinations (Orange) than M2SVid. We are visualizing only every 10th match for readability purposes. The figure … view at source ↗
Figure 11
Figure 11. Figure 11: Eye2Eye exhibits color shifts [PITH_FULL_IMAGE:figures/full_fig_p019_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Input frames, and anaglyphs generated using our [PITH_FULL_IMAGE:figures/full_fig_p020_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Comparison on Scene 158 of the AVP dataset. Note the more readable text on the bag and the texture of the ground. Note: This figure has been compressed for arXiv submission. Please refer to the project page for the uncompressed version.        3 !3  ! %$   $) + $        3 !3  ! %$   $) + $        3 !3  ! %$   $) + … view at source ↗
Figure 14
Figure 14. Figure 14: Comparison on Scene 29 of the AVP dataset. Our method preserves the branch structures in the trees. Note: This figure has been compressed for arXiv submission. Please refer to the project page for the uncompressed version. 12 [PITH_FULL_IMAGE:figures/full_fig_p023_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Comparison on Scene 42 of the AVP dataset. Our method better preserves the tree and background. Note: This figure has been compressed for arXiv submission. Please refer to the project page for the uncompressed version.        3 !3  ! %$   $) + $        3 !3  ! %$   $) + $        3 !3  ! %$   $) + $    … view at source ↗
Figure 16
Figure 16. Figure 16: Comparison on Scene 47 of the AVP dataset. Our method recovers the details on the sign. Note: This figure has been compressed for arXiv submission. Please refer to the project page for the uncompressed version. 13 [PITH_FULL_IMAGE:figures/full_fig_p024_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Comparison on Scene 121 of the AVP dataset. Our method generates a sharp output, as evidenced by the swan head. Note: This figure has been compressed for arXiv submission. Please refer to the project page for the uncompressed version.        3 !3  ! %$   $) + $        3 !3  ! %$   $) + $        3 !3  ! %$   $) + $… view at source ↗
Figure 18
Figure 18. Figure 18: Comparison on Scene 22 of the AVP dataset. Our method recovers the high-frequency details on the signs and better geometry. Note: This figure has been compressed for arXiv submission. Please refer to the project page for the uncompressed version. 14 [PITH_FULL_IMAGE:figures/full_fig_p025_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: Comparison on Scene 14 of the iPhone dataset. Our method produces readable text. Note: This figure has been compressed for arXiv submission. Please refer to the project page for the uncompressed version.        3 !3  ! %$   $) + $        3 !3  ! %$   $) + $        3 !3  ! %$   $) + $        3 … view at source ↗
Figure 20
Figure 20. Figure 20: Comparison on Scene 30 of the iPhone dataset. Our method recovers high-frequency details on the chain. Note: This figure has been compressed for arXiv submission. Please refer to the project page for the uncompressed version. 15 [PITH_FULL_IMAGE:figures/full_fig_p026_20.png] view at source ↗
Figure 21
Figure 21. Figure 21: Comparison on Scene 49 of the iPhone dataset. Our method recovers the texture on the leaves. Note: This figure has been compressed for arXiv submission. Please refer to the project page for the uncompressed version.        3 !3  ! %$   $) + $        3 !3  ! %$   $) + $        3 !3  ! %$   $) + $     … view at source ↗
Figure 22
Figure 22. Figure 22: Comparison on Scene 52 of the iPhone dataset. Our method recovers the texture on the leaves. Note: This figure has been compressed for arXiv submission. Please refer to the project page for the uncompressed version. 16 [PITH_FULL_IMAGE:figures/full_fig_p027_22.png] view at source ↗
Figure 23
Figure 23. Figure 23: Comparison on Scene 55 of the iPhone dataset. Our method preserves better the high-frequency details on text. Note: This figure has been compressed for arXiv submission. Please refer to the project page for the uncompressed version.         3 3 !  %#  #, - #         3 3 !  %#  #, - #         3 3 !  %#  #, - #       … view at source ↗
Figure 24
Figure 24. Figure 24: Comparison on Scene 56 of the iPhone dataset. Our method produces sharper details and texture. Note: This figure has been compressed for arXiv submission. Please refer to the project page for the uncompressed version. 17 [PITH_FULL_IMAGE:figures/full_fig_p028_24.png] view at source ↗
Figure 25
Figure 25. Figure 25: Anaglyphs generated by our approach on an in-the-wild image with varying disparity conditioning (in pixels). The results [PITH_FULL_IMAGE:figures/full_fig_p029_25.png] view at source ↗
Figure 26
Figure 26. Figure 26: Anaglyphs generated by our approach on an in-the-wild image with varying disparity conditioning (in pixels). The results [PITH_FULL_IMAGE:figures/full_fig_p030_26.png] view at source ↗
Figure 27
Figure 27. Figure 27: Anaglyphs generated by our approach on an in-the-wild image with varying disparity conditioning (in pixels). The results [PITH_FULL_IMAGE:figures/full_fig_p031_27.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

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

  1. UniFixer: A Universal Reference-Guided Fixer for Diffusion-Based View Synthesis

    cs.CV 2026-05 unverdicted novelty 6.0

    UniFixer is a universal reference-guided framework that fixes spatial, temporal, and backbone-related degradations in diffusion-based view synthesis via coarse-to-fine modules and achieves zero-shot SOTA results on no...

Reference graph

Works this paper leans on

72 extracted references · 11 linked inside Pith · cited by 1 Pith paper

  1. [1]

    Lumiere: A space-time diffu- sion model for video generation

    Omer Bar-Tal, Hila Chefer, Omer Tov, Charles Her- rmann, Roni Paiss, Shiran Zada, Ariel Ephrat, Junhwa Hur, Guanghui Liu, Amit Raj, et al. Lumiere: A space-time diffu- sion model for video generation. InACM SIGGRAPH Asia,

  2. [2]

    Visual competi- tion.Nature Reviews Neuroscience, 3(1):13–21, 2002

    Randolph Blake and Nikos K Logothetis. Visual competi- tion.Nature Reviews Neuroscience, 3(1):13–21, 2002. 3

    2002

  3. [3]

    Stable video diffusion: Scaling latent video diffusion models to large datasets.preprint arXiv:2311.15127, 2023

    Andreas Blattmann, Tim Dockhorn, Sumith Kulal, Daniel Mendelevitch, Maciej Kilian, Dominik Lorenz, Yam Levi, Zion English, Vikram V oleti, Adam Letts, et al. Stable video diffusion: Scaling latent video diffusion models to large datasets.preprint arXiv:2311.15127, 2023. 2, 4

    Pith/arXiv arXiv 2023

  4. [4]

    Depth Pro: Sharp monocular metric depth in less than a second.preprint arXiv:2410.02073, 2024

    Aleksei Bochkovskii, Ama ¨el Delaunoy, Hugo Germain, Marcel Santos, Yichao Zhou, Stephan R Richter, and Vladlen Koltun. Depth Pro: Sharp monocular metric depth in less than a second.preprint arXiv:2410.02073, 2024. 2

    Pith/arXiv arXiv 2024

  5. [5]

    B Breese

    B. B Breese. Binocular rivalry.Psychological Review, 16(6): 410–415, 1909. 2, 3

    1909

  6. [6]

    Com- parison of three different methods to merge multiresolution and multispectral data- landsat TM and SPOT panchromatic

    Pats Chavez, Stuart C Sides, Jeffrey A Anderson, et al. Com- parison of three different methods to merge multiresolution and multispectral data- landsat TM and SPOT panchromatic. Photogrammetric Engineering and Remote Sensing, 57(3): 295–303, 1991. 2

    1991

  7. [7]

    SVG: 3d stereoscopic video generation via denoising frame matrix

    Peng Dai, Feitong Tan, Qiangeng Xu, David Futschik, Ruofei Du, Sean Fanello, Xiaojuan Qi, and Yinda Zhang. SVG: 3d stereoscopic video generation via denoising frame matrix. InInternational Conference on Learning Represen- tations (ICLR), 2025. 1, 2, 3, 5, 6, 7

    2025

  8. [8]

    Guided super-resolution as pixel-to-pixel transformation

    Riccardo de Lutio, Stefano D’Aronco, Jan Dirk Wegner, and Konrad Schindler. Guided super-resolution as pixel-to-pixel transformation. InIEEE/CVF International Conference on Computer Vision (ICCV), 2019. 2

    2019

  9. [9]

    DeDoDe v2: Analyzing and improving the DeDoDe keypoint detec- tor

    Johan Edstedt, Georg B ¨okman, and Zhenjun Zhao. DeDoDe v2: Analyzing and improving the DeDoDe keypoint detec- tor. InIEEE/CVF Computer Society Conference on Com- puter Vision and Pattern Recognition Workshops (CVPRW),

  10. [10]

    Depth-image-based rendering (DIBR), compression, and transmission for a new approach on 3D- TV

    Christoph Fehn. Depth-image-based rendering (DIBR), compression, and transmission for a new approach on 3D- TV. InStereoscopic Displays and Virtual Reality Systems XI, pages 93–104. SPIE, 2004. 2

    2004

  11. [11]

    An image is worth one word: Personalizing text-to-image gen- eration using textual inversion.preprint arXiv:2208.01618,

    Rinon Gal, Yuval Alaluf, Yuval Atzmon, Or Patashnik, Amit H Bermano, Gal Chechik, and Daniel Cohen-Or. An image is worth one word: Personalizing text-to-image gen- eration using textual inversion.preprint arXiv:2208.01618,

    Pith/arXiv arXiv

  12. [12]

    Fine-tuning image-conditional diffusion models is easier than you think

    Gonzalo Martin Garcia, Karim Abou Zeid, Christian Schmidt, Daan De Geus, Alexander Hermans, and Bastian Leibe. Fine-tuning image-conditional diffusion models is easier than you think. InIEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2025. 4, 3

    2025

  13. [13]

    Eye2Eye: A simple approach for monocular-to-stereo video synthesis.preprint arXiv:2505.00135, 2025

    Michal Geyer, Omer Tov, Linyi Jin, Richard Tucker, Inbar Mosseri, Tali Dekel, and Noah Snavely. Eye2Eye: A simple approach for monocular-to-stereo video synthesis.preprint arXiv:2505.00135, 2025. 1, 2, 3, 5, 7, 8

    Pith/arXiv arXiv 2025

  14. [14]

    Ego4D: Around the world in 3,000 hours of egocentric video

    Kristen Grauman, Andrew Westbury, Eugene Byrne, Zachary Chavis, Antonino Furnari, Rohit Girdhar, Jack- son Hamburger, Hao Jiang, Miao Liu, Xingyu Liu, et al. Ego4D: Around the world in 3,000 hours of egocentric video. InIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022. 6, 7

    2022

  15. [15]

    Epipolar transformers

    Yihui He, Rui Yan, Katerina Fragkiadaki, and Shoou-I Yu. Epipolar transformers. InIEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), 2020. 2, 4

    2020

  16. [16]

    Denoising diffu- sion probabilistic models.Advances in Neural Information Processing Systems (NeurIPS), 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffu- sion probabilistic models.Advances in Neural Information Processing Systems (NeurIPS), 2020. 2

    2020

  17. [17]

    Video dif- fusion models.Advances in Neural Information Processing Systems (NeurIPS), 2022

    Jonathan Ho, Tim Salimans, Alexey Gritsenko, William Chan, Mohammad Norouzi, and David J Fleet. Video dif- fusion models.Advances in Neural Information Processing Systems (NeurIPS), 2022. 2

    2022

  18. [18]

    DepthCrafter: Generating consistent long depth sequences for open-world videos

    Wenbo Hu, Xiangjun Gao, Xiaoyu Li, Sijie Zhao, Xi- aodong Cun, Yong Zhang, Long Quan, and Ying Shan. DepthCrafter: Generating consistent long depth sequences for open-world videos. InIEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), pages 2005– 2015, 2025. 2, 6, 7

    2005

  19. [19]

    Restereo: Diffusion stereo video generation and restoration.preprint arXiv:2506.06023, 2025

    Xingchang Huang, Ashish Kumar Singh, Florian Dubost, Cristina Nader Vasconcelos, Sakar Khattar, Liang Shi, Chris- tian Theobalt, Cengiz Oztireli, and Gurprit Singh. Restereo: Diffusion stereo video generation and restoration.preprint arXiv:2506.06023, 2025. 1, 2, 7

    Pith/arXiv arXiv 2025

  20. [20]

    EpiDiff: Enhancing multi-view synthesis via localized epipolar-constrained diffusion

    Zehuan Huang, Hao Wen, Junting Dong, Yaohui Wang, Yangguang Li, Xinyuan Chen, Yan-Pei Cao, Ding Liang, Yu Qiao, Bo Dai, et al. EpiDiff: Enhancing multi-view synthesis via localized epipolar-constrained diffusion. InIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024. 4

    2024

  21. [21]

    Depth map super-resolution by deep multi-scale guidance

    Tak-Wai Hui, Chen Change Loy, and Xiaoou Tang. Depth map super-resolution by deep multi-scale guidance. InEu- ropean Conference on Computer Vision (ECCV), 2016. 2

    2016

  22. [22]

    Izadimehr, Milad Ghanbari, Guodong Chen, Wei Zhou, Xiaoshuai Hao, Mallesham Dasari, Christian Tim- merer, and Hadi Amirpour

    M.H. Izadimehr, Milad Ghanbari, Guodong Chen, Wei Zhou, Xiaoshuai Hao, Mallesham Dasari, Christian Tim- merer, and Hadi Amirpour. SVD: Spatial video dataset. In ACM International Conference on Multimedia (ACM MM),

  23. [23]

    Stereo4D: Learning how things move in 3d from internet stereo videos

    Linyi Jin, Richard Tucker, Zhengqi Li, David Fouhey, Noah Snavely, and Aleksander Holynski. Stereo4D: Learning how things move in 3d from internet stereo videos. InIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025. 6, 7, 8

    2025

  24. [24]

    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. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024. 2

    2024

  25. [25]

    Marigold: Affordable adaptation of diffusion- based image generators for image analysis.IEEE Transac- tions on Pattern Analysis and Machine Intelligence (TPAMI),

    Bingxin Ke, Kevin Qu, Tianfu Wang, Nando Metzger, Shengyu Huang, Bo Li, Anton Obukhov, and Konrad Schindler. Marigold: Affordable adaptation of diffusion- based image generators for image analysis.IEEE Transac- tions on Pattern Analysis and Machine Intelligence (TPAMI),

  26. [26]

    Auto-encoding varia- tional bayes

    Diederik P Kingma and Max Welling. Auto-encoding varia- tional bayes. InInternational Conference on Learning Rep- resentations (ICLR), 2013. 2

    2013

  27. [27]

    2d-to- 3d image conversion by learning depth from examples

    Janusz Konrad, Meng Wang, and Prakash Ishwar. 2d-to- 3d image conversion by learning depth from examples. In IEEE Conference on Computer Vision and Pattern Recogni- tion Workshops (CVPRW), 2012. 2

    2012

  28. [28]

    Visual discomfort of 3D TV: Assessment methods and modeling.Displays, 32(4):209–218, 2011

    Marc Lambooij, Wijnand A IJsselsteijn, and Ingrid Heynd- erickx. Visual discomfort of 3D TV: Assessment methods and modeling.Displays, 32(4):209–218, 2011. 3

    2011

  29. [29]

    Stereo conversion with disparity-aware warping, compositing and inpainting

    Lukas Mehl, Andr ´es Bruhn, Markus Gross, and Christopher Schroers. Stereo conversion with disparity-aware warping, compositing and inpainting. InIEEE/CVF Winter Confer- ence on Applications of Computer Vision (WACV), 2024. 3

    2024

  30. [30]

    Guided depth super-resolution by deep anisotropic diffusion

    Nando Metzger, Rodrigo Caye Daudt, and Konrad Schindler. Guided depth super-resolution by deep anisotropic diffusion. InIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023. 2

    2023

  31. [31]

    T2I-Adapter: Learn- ing adapters to dig out more controllable ability for text-to- image diffusion models

    Chong Mou, Xintao Wang, Liangbin Xie, Yanze Wu, Jian Zhang, Zhongang Qi, and Ying Shan. T2I-Adapter: Learn- ing adapters to dig out more controllable ability for text-to- image diffusion models. InAAAI Conference on Artificial Intelligence, 2024. 3

    2024

  32. [32]

    One-step image translation with text-to-image models.preprint arXiv:2403.12036, 2024

    Gaurav Parmar, Taesung Park, Srinivasa Narasimhan, and Jun-Yan Zhu. One-step image translation with text-to-image models.preprint arXiv:2403.12036, 2024. 2

    Pith/arXiv arXiv 2024

  33. [33]

    UniDepth: Universal monocular metric depth estimation

    Luigi Piccinelli, Yung-Hsu Yang, Christos Sakaridis, Mattia Segu, Siyuan Li, Luc Van Gool, and Fisher Yu. UniDepth: Universal monocular metric depth estimation. InIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024. 2

    2024

  34. [34]

    UniDepthV2: Universal monocular metric depth estimation made simpler.IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2025

    Luigi Piccinelli, Christos Sakaridis, Yung-Hsu Yang, Mat- tia Segu, Siyuan Li, Wim Abbeloos, and Luc Van Gool. UniDepthV2: Universal monocular metric depth estimation made simpler.IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2025. 2

    2025

  35. [35]

    SDXL: Improving latent diffusion models for high-resolution image synthesis

    Dustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas M ¨uller, Joe Penna, and Robin Rombach. SDXL: Improving latent diffusion models for high-resolution image synthesis. InInternational Con- ference on Learning Representations (ICLR), 2024. 2

    2024

  36. [36]

    Towards open-world generation of stereo images and unsu- pervised matching

    Feng Qiao, Zhexiao Xiong, Eric Xing, and Nathan Jacobs. Towards open-world generation of stereo images and unsu- pervised matching. InIEEE/CVF International Conference on Computer Vision (ICCV), 2025. 2, 5

    2025

  37. [37]

    Ren ´e Ranftl, Katrin Lasinger, David Hafner, Konrad Schindler, and Vladlen Koltun. Towards robust monocular depth estimation: Mixing datasets for zero-shot cross-dataset transfer.IEEE Transactions on Pattern Analysis and Ma- chine Intelligence (TPAMI), 44(3):1623–1637, 2020. 2

    2020

  38. [38]

    Vi- sion transformers for dense prediction

    Ren ´e Ranftl, Alexey Bochkovskiy, and Vladlen Koltun. Vi- sion transformers for dense prediction. InIEEE/CVF In- ternational Conference on Computer Vision (ICCV), pages 12179–12188, 2021. 2

    2021

  39. [39]

    Deepspeed: System optimizations enable train- ing deep learning models with over 100 billion parameters

    Jeff Rasley, Samyam Rajbhandari, Olatunji Ruwase, and Yuxiong He. Deepspeed: System optimizations enable train- ing deep learning models with over 100 billion parameters. InProceedings of the 26th ACM SIGKDD international con- ference on knowledge discovery & data mining, pages 3505– 3506, 2020. 1

    2020

  40. [40]

    High-resolution image syn- thesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj¨orn Ommer. High-resolution image syn- thesis with latent diffusion models. InIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022. 2

    2022

  41. [41]

    U- net: Convolutional networks for biomedical image segmen- tation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U- net: Convolutional networks for biomedical image segmen- tation. InMedical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18, pages 234–241. Springer, 2015. 3

    2015

  42. [42]

    DreamBooth: Fine tuning text-to-image diffusion models for subject-driven generation

    Nataniel Ruiz, Yuanzhen Li, Varun Jampani, Yael Pritch, Michael Rubinstein, and Kfir Aberman. DreamBooth: Fine tuning text-to-image diffusion models for subject-driven generation. InIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023. 3

    2023

  43. [43]

    Photorealistic text-to-image diffusion models with deep language understanding.Advances in Neural Information Processing Systems (NeurIPS), 2022

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily L Denton, Kamyar Ghasemipour, Raphael Gontijo Lopes, Burcu Karagol Ayan, Tim Salimans, et al. Photorealistic text-to-image diffusion models with deep language understanding.Advances in Neural Information Processing Systems (NeurIPS), 2022. 2

    2022

  44. [44]

    Progressive distillation for fast sampling of diffusion models.arXiv preprint arXiv:2202.00512, 2022

    Tim Salimans and Jonathan Ho. Progressive distillation for fast sampling of diffusion models.arXiv preprint arXiv:2202.00512, 2022. 3

    Pith/arXiv arXiv 2022

  45. [45]

    Feng Shao, Weisi Lin, Shanbo Gu, Gangyi Jiang, and Tham- bipillai Srikanthan. Perceptual full-reference quality assess- ment of stereoscopic images by considering binocular vi- sual characteristics.IEEE Transactions on Image Processing (TIP), 22(5):1940–1953, 2013. 3

    1940

  46. [46]

    ImmersePro: End- to-end stereo video synthesis via implicit disparity learning

    Jian Shi, Zhenyu Li, and Peter Wonka. ImmersePro: End- to-end stereo video synthesis via implicit disparity learning. preprint arXiv:2410.00262, 2024. 3

    Pith/arXiv arXiv 2024

  47. [47]

    The zone of comfort: Predicting visual discom- fort with stereo displays.Journal of Vision, 11(8):11–11,

    Takashi Shibata, Joohwan Kim, David M Hoffman, and Mar- tin S Banks. The zone of comfort: Predicting visual discom- fort with stereo displays.Journal of Vision, 11(8):11–11,

  48. [48]

    3d photography using context-aware layered depth inpainting

    Meng-Li Shih, Shih-Yang Su, Johannes Kopf, and Jia-Bin Huang. 3d photography using context-aware layered depth inpainting. InIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020. 3

    2020

  49. [49]

    3d scene reconstruction with multi-layer depth and epipolar transformers

    Daeyun Shin, Zhile Ren, Erik B Sudderth, and Charless C Fowlkes. 3d scene reconstruction with multi-layer depth and epipolar transformers. InIEEE/CVF International Confer- ence on Computer Vision (ICCV), 2019. 2

    2019

  50. [50]

    M2SVid: End-to-end inpainting and refinement for monocular-to-stereo video conversion

    Nina Shvetsova, Goutam Bhat, Prune Truong, Hilde Kuehne, and Federico Tombari. M2SVid: End-to-end inpainting and refinement for monocular-to-stereo video conversion. InIn- ternational Conference on 3D Vision (3DV), 2026. 1, 2, 3, 4, 5, 6, 7, 10

    2026

  51. [51]

    Denois- ing diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denois- ing diffusion implicit models. InInternational Conference on Learning Representations (ICLR), 2021. 2

    2021

  52. [52]

    What makes for a good stereoscopic image? InIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025

    Netanel Tamir, Shir Amir, Ranel Itzhaky, Noam Atia, Shob- hita Sundaram, Stephanie Fu, Ron Sokolovsky, Phillip Isola, 10 Tali Dekel, Richard Zhang, et al. What makes for a good stereoscopic image? InIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025. 7

    2025

  53. [53]

    RAFT: Recurrent all-pairs field transforms for optical flow

    Zachary Teed and Jia Deng. RAFT: Recurrent all-pairs field transforms for optical flow. InEuropean Conference on Computer Vision (ECCV), 2020. 6

    2020

  54. [54]

    Methods for reducing vi- sual discomfort in stereoscopic 3d: A review.Signal Pro- cessing: Image Communication, 47:402–416, 2016

    Kasim Terzi ´c and Miles Hansard. Methods for reducing vi- sual discomfort in stereoscopic 3d: A review.Signal Pro- cessing: Image Communication, 47:402–416, 2016. 3

    2016

  55. [55]

    Attention is all you need.Advances in neural information processing systems, 30, 2017

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszko- reit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need.Advances in neural information processing systems, 30, 2017. 2

    2017

  56. [56]

    A critical comparison of pansharpening algorithms

    Gemine Vivone, Luciano Alparone, Jocelyn Chanussot, Mauro Dalla Mura, Andrea Garzelli, Giorgio Licciardi, Rocco Restaino, and Lucien Wald. A critical comparison of pansharpening algorithms. InIEEE Geoscience and Remote Sensing Symposium (IGARSS), 2014. 2

    2014

  57. [57]

    StereoDiffusion: Training- free stereo image generation using latent diffusion models

    Lezhong Wang, Jeppe Revall Frisvad, Mark Bo Jensen, and Siavash Arjomand Bigdeli. StereoDiffusion: Training- free stereo image generation using latent diffusion models. InIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024. 2

    2024

  58. [58]

    Image quality assessment: from error visibility to structural similarity.IEEE Transactions on Image Process- ing (TIP), 13(4):600–612, 2004

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Si- moncelli. Image quality assessment: from error visibility to structural similarity.IEEE Transactions on Image Process- ing (TIP), 13(4):600–612, 2004. 5

    2004

  59. [59]

    Learning stereo from single images

    Jamie Watson, Oisin Mac Aodha, Daniyar Turmukhambe- tov, Gabriel J Brostow, and Michael Firman. Learning stereo from single images. InEuropean Conference on Computer Vision (ECCV), 2020. 3

    2020

  60. [60]

    FoundationStereo: Zero- shot stereo matching.IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025

    Bowen Wen, Matthew Trepte, Joseph Aribido, Jan Kautz, Orazio Gallo, and Stan Birchfield. FoundationStereo: Zero- shot stereo matching.IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025. 6

    2025

  61. [61]

    Direct and explicit 3d generation from a single image

    Haoyu Wu, Meher Gitika Karumuri, Chuhang Zou, Seung- bae Bang, Yuelong Li, Dimitris Samaras, and Sunil Hadap. Direct and explicit 3d generation from a single image. In International Conference on 3D Vision (3DV), 2025. 3

    2025

  62. [62]

    Deep3d: Fully automatic 2d-to-3d video conversion with deep convo- lutional neural networks

    Junyuan Xie, Ross Girshick, and Ali Farhadi. Deep3d: Fully automatic 2d-to-3d video conversion with deep convo- lutional neural networks. InEuropean Conference on Com- puter Vision (ECCV), 2016. 2

    2016

  63. [63]

    Depth Anything: Unleash- ing the power of large-scale unlabeled data

    Lihe Yang, Bingyi Kang, Zilong Huang, Xiaogang Xu, Jiashi Feng, and Hengshuang Zhao. Depth Anything: Unleash- ing the power of large-scale unlabeled data. InIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024. 2

    2024

  64. [64]

    Any-to-Bokeh: One-step video bokeh via multi-plane image guided diffu- sion.preprint arXiv:2505.21593, 2025

    Yang Yang, Siming Zheng, Jinwei Chen, Boxi Wu, Xiaofei He, Deng Cai, Bo Li, and Peng-Tao Jiang. Any-to-Bokeh: One-step video bokeh via multi-plane image guided diffu- sion.preprint arXiv:2505.21593, 2025. 2

    arXiv 2025

  65. [65]

    IP- Adapter: Text compatible image prompt adapter for text-to- image diffusion models.preprint arXiv:2308.06721, 2023

    Hu Ye, Jun Zhang, Sibo Liu, Xiao Han, and Wei Yang. IP- Adapter: Text compatible image prompt adapter for text-to- image diffusion models.preprint arXiv:2308.06721, 2023. 3

    Pith/arXiv arXiv 2023

  66. [66]

    Tra- jectoryCrafter: Redirecting camera trajectory for monocu- lar videos via diffusion models

    Mark YU, Wenbo Hu, Jinbo Xing, and Ying Shan. Tra- jectoryCrafter: Redirecting camera trajectory for monocu- lar videos via diffusion models. InIEEE/CVF International Conference on Cputer Vision (ICCV), 2025. 3

    2025

  67. [67]

    3D- TV content creation: automatic 2D-to-3D video conversion

    Liang Zhang, Carlos Vazquez, and Sebastian Knorr. 3D- TV content creation: automatic 2D-to-3D video conversion. IEEE Transactions on Broadcasting, 57(2):372–383, 2011. 2

    2011

  68. [68]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. In IEEE/CVF International Conference on Computer Vision (ICCV), 2023. 3, 2

    2023

  69. [69]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shecht- man, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. InIEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018. 5

    2018

  70. [70]

    High-fidelity novel view synthesis via splatting-guided diffusion

    Xiang Zhang, Yang Zhang, Lukas Mehl, Markus Gross, and Christopher Schroers. High-fidelity novel view synthesis via splatting-guided diffusion. InACM SIGGRAPH, 2025. 2, 5

    2025

  71. [71]

    StereoCrafter: Diffusion-based generation of long and high- fidelity stereoscopic 3d from monocular videos.preprint arXiv:2409.07447, 2024

    Sijie Zhao, Wenbo Hu, Xiaodong Cun, Yong Zhang, Xiaoyu Li, Zhe Kong, Xiangjun Gao, Muyao Niu, and Ying Shan. StereoCrafter: Diffusion-based generation of long and high- fidelity stereoscopic 3d from monocular videos.preprint arXiv:2409.07447, 2024. 1, 2, 3, 4, 6, 7, 10

    Pith/arXiv arXiv 2024

  72. [72]

    structurally

    Wei Zhou, Hadi Amirpour, Christian Timmerer, Guangtao Zhai, Patrick Le Callet, and Alan C Bovik. Perceptual visual quality assessment: Principles, methods, and future direc- tions.preprint arXiv:2503.00625, 2025. 3 11 Elastic3D: Controllable Stereo Video Conversion with Guided Latent Decoding Supplementary Material In Sec. A, we provide implementation det...

    Pith/arXiv arXiv 2025