Pith. sign in

REVIEW 3 major objections 4 minor 51 references

Object-Driven Multi-Layer Scene Decomposition From a Single Image

T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read A single RGB image can be decomposed into layered color and depth that complete occluded objects, and the paper shows this beats prior LDI and view-synthesis baselines on indoor scenes.

desk verdict A useful, well-ablated extension of single-image LDI prediction, but the re-composition loss in Eq. (8) is printed wrong and must be fixed before the paper is reproducible. read the letter →

arxiv 1908.09521 v1 pith:X37ZL4UH submitted 2019-08-26 cs.CV

Helisa Dhamo , Nassir Navab , Federico Tombari This is my paper · ORCID
classification cs.CV
keywords layereddepthimagesingle-imagescenedecompositionoccludedobjectcompletionpredictionviewsynthesisremovaldiminishedrealitysemanticunderstanding
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that a single color image contains enough information to reconstruct a multi-layer depth representation of the scene, including the appearance and depth of objects hidden behind other objects. The authors build a Layered Depth Image (LDI) whose layers are generated object by object, with the number of layers adapting to the scene rather than being fixed in advance. They argue that adding semantic class information and a re-composition loss improves the hallucination of occluded content. If correct, this would make single-image 3D photography, view synthesis, and object removal more practical and accurate than previous two-layer approaches.

What carries the argument

The load-bearing mechanism is a three-network pipeline: an object completion encoder-decoder that takes the RGB image, predicted mask, and class scores and outputs a completed RGBA-D representation of each object; a U-Net layout generator that predicts a background scene layout with adversarial, perceptual, and reconstruction losses; and a minimum depth pooling (MDP) re-composition block that concatenates predicted layers, selects per-pixel the layer with smallest depth, and learns depth displacements to align visible regions with the input depth. The MDP re-composition loss is what ties the independently predicted object and layout layers into a coherent depth-ordered scene.

What would settle it

Run the full pipeline on a real indoor dataset with complete, hole-free meshes and compare second-layer depth and color errors against the paper's reported numbers; if the gains over prior baselines shrink substantially, the reported advantage is partly an artifact of incomplete rendering targets. A simpler check is to replace Mask R-CNN predictions with ground-truth instance masks at test time and measure whether the second-layer accuracy changes.

Watch

Extended reading notes

Core claim

The paper claims that an object-driven decomposition with adaptive layer count outperforms prior methods for LDI prediction from a single image, and that the resulting layered representation yields better view synthesis. The method detects objects with Mask R-CNN, completes each object's color, transparency, and depth into an RGBA-D map, predicts an object-free scene layout, and then recomposes all layers using a minimum depth pooling step that enforces global depth consistency. On SunCG, the reported second-layer depth error is lower than both prior baselines, and view synthesis SSIM is higher; results on Stanford 2D-3D confirm the second-layer gains in a real-world setting.

Load-bearing premise

The method's training targets are generated by rendering scene meshes, and the authors note that real datasets like Stanford 2D-3D contain holes and missing surfaces, so the ground-truth occluded layers may themselves be incomplete or biased.

Editorial extensions

If this is right

  • A single RGB image can produce more than two depth layers, so occluded intermediate objects appear in the layered output rather than being merged into background.
  • Adding semantic class scores to object completion improves occluded-region depth and color accuracy compared to class-agnostic completion.
  • Enforcing recomposed-image consistency through minimum depth pooling improves depth accuracy not only for visible object regions but also for occluded parts.
  • The object-wise representation directly supports object removal and diminished reality, since a requested category can be dropped from the layers and the remaining layers re-render the scene.
  • Better LDI prediction translates to higher view-synthesis quality under simple image-based rendering, measured by SSIM, MPE, and RMSE.

Reading between the lines

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

  • If the reported gains hold, the same object-driven decomposition could be applied to outdoor scenes by swapping the object detector and layout prior, since the method itself is not indoor-specific.
  • The reliance on Mask R-CNN suggests a testable extension: replacing predicted masks with ground-truth masks should isolate how much of the second-layer accuracy depends on detector quality versus the completion networks.
  • A future dataset with hole-free real-scene meshes could resolve whether the Stanford 2D-3D results understate the method's true capability or are partly inflated by the renderer's missing surfaces.
  • The adaptive layer count implies the method's error should be reported per-layer frequency, which the paper does in the supplement; a natural extension is a practical heuristic for choosing the number of layers at inference time based on detection confidence.
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 / 4 minor

Summary. The paper proposes an object-driven approach to single-image Layered Depth Image (LDI) prediction. It uses Mask R-CNN to obtain instance masks and class scores, completes each partially occluded object's RGBA-D with Network A, predicts a scene layout with Network B, and assembles the layers through a minimum-depth-pooling (MDP) re-composition step supervised by Network C. Supervision is generated by mesh-based rendering from SunCG and Stanford 2D-3D. The method is evaluated against Dhamo et al. [6] and Tulsiani et al. [41] on LDI prediction, view synthesis, and object removal, reporting improvements on SunCG and on occluded-layer structure on Stanford 2D-3D. The paper also introduces two rendered datasets for the task.

Significance. If the technical issues are resolved, the paper would be a solid contribution: it extends LDI prediction from fixed two-layer models to an object-adaptive number of layers, demonstrates that semantic and instance cues help complete occluded object regions, and provides a mesh-rendered supervision pipeline with two datasets for future work. The reported gains over [6] and [41] are large in the SunCG comparisons, e.g., Table 1 second-layer depth MPE of 0.641 versus 1.139 and 1.582, and Table 3 SSIM of 0.65 versus 0.56 and 0.33. The paper is generally well organized, with additional ablations, qualitative results, and failure cases in the supplement. Its main weakness is that the central re-composition loss is not defined reproducibly, and a few component-ablation claims are not fully supported by the reported numbers.

major comments (3)
  1. [Sec. 3.4, Eqs. (7)-(8)] The re-composition loss as printed is not the loss the text and Fig. 2 describe. Eq. (7) defines delta_l as a scalar per layer (the difference between mean ground-truth depth and mean predicted depth over the binary mask m_l), and Eq. (8) computes || y_{delta,l} - hat_y_{d,l} ||_1 with y_{delta,l} = hat_y_{d,l} + delta_l, which reduces to || delta_l ||_1. This is a per-layer mean-shift regularizer; it neither compares the displaced depth to the ground-truth depth y_{d,l} nor enforces that the recomposed layers encode the structure of the original input, contrary to the claim in Sec. 3.4 and the Fig. 2 caption. Consequently, the '+Lrecompose' row in Table 1 is not reproducible from the paper. Please correct Eq. (8), most plausibly to L_recompose = || y_{d,l} - (hat_y_{d,l} + delta_l) ||_1, and clarify whether delta_l is a learned per-pixel displacement map or a deterministic scalar. The corrected definition should also be reflected in Fig. 2 and in the ablation description.
  2. [Sec. 3.4, MDP layer] The minimum-depth-pooling operation selects, for each pixel, the layer with the lowest predicted depth, producing an index map via an argmin. This selection is non-differentiable, and the paper does not describe how gradients from L_recompose propagate through the layer-order selection to the depth values that determine the ordering. As a result, the claim that the network 'implicitly and globally enforce[s] the depth consistency' of layer sorting is not established. Please provide a detailed treatment of the MDP layer's gradient flow (e.g., a straight-through estimator or a detached mask) and state explicitly whether the layer ordering is supervised during training.
  3. [Sec. 5.1, Table 1] The ablation claims in the text are not fully supported by the reported numbers. From the 'Ours, baseline (w/o class scores)' row to the '+ class scores' row, the second-layer depth MPE increases from 0.687 to 0.700, while the first-layer MPE improves from 0.551 to 0.508. The sentence stating 'one can observe an improvement from adding the class category component ... especially for depth' is therefore inaccurate for the second layer. In addition, no error bars or significance tests are provided, so the small differences between ablation rows (e.g., second-layer depth MPE 0.657 versus 0.641 for +Lp versus +Lrecompose) cannot be distinguished from noise. Please report per-component results accurately, include uncertainty estimates or significance testing, and temper the component-wise improvement claims accordingly.
minor comments (4)
  1. [Sec. 3.4] There is a typo in the sentence beginning 'Foe each layer l' near Eq. (7); it should read 'For each layer l'.
  2. [Sec. 4, zero-padding paragraph] The sentence 'the the original view is spanned' contains a duplicated article; it should be 'the original view'.
  3. [Sec. 3.3, Eq. (6)] The notation for y_c and y_d is inconsistent: Eq. (4) treats them as ground-truth layouts, but the sentence following Eq. (6) says 'where y_c, y_d denote output color and depth respectively.' Please unify the notation to distinguish ground truth from predictions.
  4. [General] The paper states that the two datasets will be publicly released, but code availability is not mentioned. Given the ambiguity in the re-composition loss definition, releasing code (or at least pseudocode for the MDP and L_recompose computation) would materially aid reproducibility.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: predictions are supervised from rendered ground truth and benchmarked against external baselines; the only self-citation is a non-load-bearing baseline.

full rationale

The paper's load-bearing claim is that a single RGB image can be decomposed into an object-driven Layered Depth Image with completed occluded layers, outperforming prior work. This is established by supervised training on automatically rendered ground-truth RGBA-D layers (Section 3.1, Eqs. (1)-(6)) and by comparison with Tulsiani et al. [41] and Dhamo et al. [6] on SunCG and Stanford 2D-3D (Tables 1-3). The network outputs are not defined in terms of the evaluation metrics or the test inputs: the targets are obtained from explicit mesh rendering of the scene, while the inputs are RGB, Mask R-CNN mask/class predictions, and a monocular depth prior. No fitted parameter is renamed as a prediction, and no uniqueness theorem or ansatz is imported from the authors' prior work. The only overlapping-author reference is [6], which appears as an independently published baseline and as a data-generation contrast ('Unlike Dhamo et al. [6], we employ a mesh-based rendering approach'); this citation is not load-bearing for the claimed derivation. One internal inconsistency is noted for completeness but is not circularity: substituting Eq. (7) into Eq. (8) gives L_recompose = ||δ_l||_1, so the printed loss only aligns the per-layer mean depth with the ground-truth mean rather than enforcing the 'image re-composition' described in the text and Fig. 2; this is a reproducibility/correctness concern. The acknowledged limitations (mesh holes in Stanford 2D-3D, undetected objects) are data-quality issues scoped to the real-data experiment and do not make the evaluation circular. The derivation chain is therefore self-contained, with at most a minor, non-load-bearing self-citation.

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

The central claims rest on conventional supervised deep learning assumptions, on the quality of automatically rendered layered ground truth, and on the reliability of upstream Mask R-CNN and depth-prediction modules. The main free parameters are hand-set loss weights and thresholds that are not independently justified; they affect quantitative outcomes but do not by themselves force the reported comparisons.

free parameters (4)
  • Relevance map weights gamma in L_completion = 0.7 in visible region, 1.5 in occluded region, 0.2 otherwise
    Chosen by hand in Section 3.2 to weight the completion L1 loss; no sensitivity analysis reported.
  • Layout loss weights lambda_r and lambda_p = lambda_r = 100, lambda_p = 25
    Set in Section 4 for the adversarial layout objective (Eq.3); no ablation over these weights.
  • Mask matching IoU threshold = 0.3
    Used in Section 3.2 to pair Mask R-CNN predictions with ground-truth masks; discards detections with IoU below the threshold and shapes the training distribution.
  • Zero-padding band sizes = 16 pixels top/bottom, 12 pixels left/right
    Set in Section 4 so novel views contain border content; arbitrary and dataset-specific.
assumptions (4)
  • domain assumption Mesh-rendered ground truth layers are faithful to real occlusion and disocclusion.
    Section 3.1 builds all supervision from SunCG and Stanford 2D-3D meshes; the authors note Stanford meshes have holes, so the assumption is only partially met.
  • domain assumption Mask R-CNN detections and class scores are good enough at test time.
    Network A runs only on predicted masks; the supplement failure cases state missed detections are dropped from layers.
  • domain assumption The external monocular depth prior (Laina et al. [23]) provides a reliable global depth reference.
    Network B uses this depth as input, and MDP uses masked predicted depth to define front regions; wrong prior depth breaks layer ordering.
  • standard math Conventional backpropagation with Adam optimizes the stated losses.
    The paper assumes standard deep-learning training behavior; no convergence guarantee is needed for the empirical claims.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Object-Driven Multi-Layer Scene Decomposition From a Single Image." pith.science (2026). https://pith.science/paper/X37ZL4UH

@misc{pith2026190809521,
  author       = {Pith},
  title        = {Pith review of: Object-Driven Multi-Layer Scene Decomposition From a Single Image},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/X37ZL4UH}},
  note         = {Machine review of arXiv:1908.09521}
}
read the original abstract

We present a method that tackles the challenge of predicting color and depth behind the visible content of an image. Our approach aims at building up a Layered Depth Image (LDI) from a single RGB input, which is an efficient representation that arranges the scene in layers, including originally occluded regions. Unlike previous work, we enable an adaptive scheme for the number of layers and incorporate semantic encoding for better hallucination of partly occluded objects. Additionally, our approach is object-driven, which especially boosts the accuracy for the occluded intermediate objects. The framework consists of two steps. First, we individually complete each object in terms of color and depth, while estimating the scene layout. Second, we rebuild the scene based on the regressed layers and enforce the recomposed image to resemble the structure of the original input. The learned representation enables various applications, such as 3D photography and diminished reality, all from a single RGB image.

Figures

Figures reproduced from arXiv: 1908.09521 by the authors.

Figure 1
Figure 1. Overview of our method and its applications. (Top): Given a single color image, we infer a layered representation that consists of a set of RGB and depth images for every object in the scene, as well as the scene layout. (Bottom): Illustration of two applications, a) view synthesis and b) object removal. was originally taken from, the visual content around object borders gets dis-occluded, thus enabling the image fr… view at source ↗
Figure 2
Figure 2. Proposed scene layering framework. Left: While Network A (top) completes the occluded parts for each detected object instance, to a RGBA-D representation, Network B (bottom) predicts RGBA and depth images for the empty scene. Right: The outputs are concatenated and fed to the Minimum Depth Pooling (MDP) layer, that recomposes the scene. Instance-wise, the displacement of the recomposed first layer depth from the gro… view at source ↗
Figure 3
Figure 3. Overview of the proposed object completion encoder architecture. Class probabilities branch (left) and image branch (right) are concatenated along channels in the bottleneck layer. where x is the ground truth RGBA-D map and xˆ the auto￾encoded output. For this unsupervised learning part, we use those object instances that do not touch the image borders to guarantee visibility on the whole object appearance. Then, we… view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: LDI prediction results on SunCG. Left: The input color image. Center: Our predictions for the first two layers, obtained after sorting the object-wise layers. Right: Ground truth, as extracted from the mesh-based rendering. Method 1st layer depth 2nd layer depth 2nd la…
Figure 5
Figure 5. Figure 5: LDI prediction results on Stanford 2D-3D. Left: The input color image. Center: Our predictions for the first two layers, obtained after sorting the object-wise layers. Right: Ground truth, as extracted from the mesh-based rendering. Black in the color images and dark b…
Figure 6
Figure 6. Figure 6: Illustration of object removal results. The category labels of the left indicate which object should be removed from the original image. We compare our predicted synthesized images (center) against the ground truth (right). appears sparse, although as mentioned in Sect…
Figure 7
Figure 7. Figure 7: illustrates how our method preserves the shapes of the front objects while rendering [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: LDI prediction results on SunCG. For each example, Left: The input color image. Right: From top to bottom - ground truth, two-layer predictions of the proposed method, Dhamo et al. [6] and Tulsiani et al. [41] for the first two layers. SunCG Stanford 2D-3D [PITH_FULL_…
Figure 9
Figure 9. Figure 9: Multi-layer evaluation for SunCG (top) and Stanford 2D-3D (bottom). Left: The layer frequency, i.e. for layer l the frequency of images that have an l th layer. Center: Color MPE and RMSE errors. Right: Depth MPE and RMSE errors. 2 [PITH_FULL_IMAGE:figures/full_fig_p0…
Figure 10
Figure 10. Figure 10: RGBA object completion and layout prediction results on Stanford 2D-3D. Input image, instance examples (top to bottom: mask, prediction, ground truth) as well as layout prediction. 3 [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]
Figure 11
Figure 11. Figure 11: RGBA object completion and layout prediction results on SunCG. Input image, instance examples (top to bottom: mask, prediction, ground truth) as well as layout prediction. 4 [PITH_FULL_IMAGE:figures/full_fig_p014_11.png]
Figure 12
Figure 12. Figure 12: Visualization of the visible masks. Left: SunCG, Right: Stanford 2D-3D. Top: Instance masks as predicted from Mask R-CNN, i.e. input to our object completion network. Bottom: Instance masks using the visible parts of our predicted object extent. We observe that the ob…
Figure 13
Figure 13. Figure 13: Illustration of the SunCG dataset. For every view, we provide the RGBA, depth, instance segmentation and class categories. This applies for the full-image content, object-wise layers as well as the layout. Even though the layout components are merged into a single lay…
Figure 14
Figure 14. Figure 14: Illustration of the Stanford 2D-3D dataset. For every view, we provide the RGBA, depth, instance segmentation and class categories. This applies for the full-image content, object-wise layers as well as the layout. Even though the layout components are merged into a s…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 43 canonical work pages

  1. [6]

    Peeking behind objects: Layered depth prediction from a single image

    Helisa Dhamo, Keisuke Tateno, Iro Laina, Nassir Navab, and Federico Tombari. Peeking behind objects: Layered depth prediction from a single image. Pattern Recognition Letters, 2019

    work page 2019

  2. [41]

    Layer-structured 3d scene inference via view synthesis

    Shubham Tulsiani, Richard Tucker, and Noah Snavely. Layer-structured 3d scene inference via view synthesis. In ECCV, 2018

    work page 2018

  3. [1]

    Joint 2d-3d-semantic data for indoor scene under- standing

    Iro Armeni, Sasha Sax, Amir Roshan Zamir, and Silvio Savarese. Joint 2d-3d-semantic data for indoor scene under- standing. arXiv:1702.01105, 2017

    arXiv 2017

  4. [2]

    Jimmy Ba, Jamie Ryan Kiros, and Geoffrey E. Hinton. Layer normalization. arXiv:1607.06450, 2016

    arXiv 2016

  5. [3]

    Simon Baker, Richard Szeliski, and P. Anandan. A layered approach to stereo reconstruction. CVPR, 1998

    work page 1998

  6. [4]

    3d-r2n2: A unified approach for single and multi-view 3d object reconstruction

    Christopher B Choy, Danfei Xu, JunYoung Gwak, Kevin Chen, and Silvio Savarese. 3d-r2n2: A unified approach for single and multi-view 3d object reconstruction. In ECCV, 2016

    work page 2016

  7. [5]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kehui Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. CVPR, 2009

    work page 2009

  8. [7]

    Segan: Segmenting and generating the invisible

    Kiana Ehsani, Roozbeh Mottaghi, and Ali Farhadi. Segan: Segmenting and generating the invisible. In CVPR, 2018

    work page 2018

Show all 51 references
  1. [8]

    Depth map prediction from a single image using a multi-scale deep net- work

    David Eigen, Christian Puhrsch, and Rob Fergus. Depth map prediction from a single image using a multi-scale deep net- work. In NIPS, 2014

    2014

  2. [9]

    Haoqiang Fan, Hao Su, and Leonidas J. Guibas. A point set generation network for 3d object reconstruction from a single image. In CVPR, 2017

    2017

  3. [10]

    Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio

    Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. In NIPS, 2014

    2014

  4. [11]

    Predicting complete 3d models of indoor scenes

    Ruiqi Guo, Chuhang Zou, and Derek Hoiem. Predicting complete 3d models of indoor scenes. arXiv:1504.02437, 2015

    2015 arXiv

  5. [12]

    Girshick

    Kaiming He, Georgia Gkioxari, Piotr Doll ´ar, and Ross B. Girshick. Mask r-cnn. ICCV, 2017

    2017

  6. [13]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. CVPR, 2016

    2016

  7. [14]

    Casual 3d photography

    Peter Hedman, Suhib Alsisan, Richard Szeliski, and Jo- hannes Kopf. Casual 3d photography. ACM Trans. Graph., 2017

    2017

  8. [15]

    Densely connected convolutional net- works

    Gao Huang, Zhuang Liu, Laurens van der Maaten, and Kil- ian Q Weinberger. Densely connected convolutional net- works. In CVPR, 2017

    2017

  9. [16]

    Scene collaging: Analysis and syn- thesis of natural images with semantic layers

    Phillip Isola and Ce Liu. Scene collaging: Analysis and syn- thesis of natural images with semantic layers. ICCV, 2013

    2013

  10. [17]

    Image-to-image translation with conditional adver- sarial networks

    Phillip Isola, Jun-Yan Zhu, Tinghui Zhou, and Alexei A Efros. Image-to-image translation with conditional adver- sarial networks. ICCV, 2017

    2017

  11. [18]

    Jianbo Jiao, Ying Cao, Yibing Song, and Rynson W. H. Lau. Look deeper into depth: Monocular depth estimation with semantic booster and attention-driven loss. In ECCV, 2018

    2018

  12. [19]

    Perceptual losses for real-time style transfer and super-resolution

    Justin Johnson, Alexandre Alahi, and Li Fei-Fei. Perceptual losses for real-time style transfer and super-resolution. In ECCV, 2016

    2016

  13. [20]

    Depth trans- fer: Depth extraction from videos using nonparametric sam- pling

    Kevin Karsch, Ce Liu, and Sing Bing Kang. Depth trans- fer: Depth extraction from videos using nonparametric sam- pling. In Dense Image Correspondences for Computer Vi- sion. 2016

    2016

  14. [21]

    What uncertainties do we need in bayesian deep learning for computer vision? In NIPS, 2017

    Alex Kendall and Yarin Gal. What uncertainties do we need in bayesian deep learning for computer vision? In NIPS, 2017

    2017

  15. [22]

    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. InCVPR Workshops, 2012

    2012

  16. [23]

    Deeper depth prediction with fully convolutional residual networks

    Iro Laina, Christian Rupprecht, Vasileios Belagiannis, Fed- erico Tombari, and Nassir Navab. Deeper depth prediction with fully convolutional residual networks. In 3DV, 2016

    2016

  17. [24]

    van den Hengel, and Mingyi He

    Bo Li, Chunhua Shen, Yuchao Dai, A. van den Hengel, and Mingyi He. Depth and surface normal estimation from monocular images using regression on deep features and hi- erarchical crfs. In CVPR, 2015

    2015

  18. [25]

    Microsoft coco: Common objects in context

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll´ar, and C Lawrence Zitnick. Microsoft coco: Common objects in context. In ECCV, 2014

    2014

  19. [26]

    Single im- age depth estimation from predicted semantic labels

    Beyang Liu, Stephen Gould, and Daphne Koller. Single im- age depth estimation from predicted semantic labels. CVPR, 2010

    2010

  20. [27]

    Planercnn: 3d plane detection and reconstruction from a single image

    Chen Liu, Kihwan Kim, Jinwei Gu, Yasutaka Furukawa, and Jan Kautz. Planercnn: 3d plane detection and reconstruction from a single image. In CVPR, 2019

    2019

  21. [28]

    Layered scene decomposition via the occlusion-crf

    Chen Liu, Pushmeet Kohli, and Yasutaka Fukurawa. Layered scene decomposition via the occlusion-crf. In CVPR, 2016

    2016

  22. [29]

    Deep con- volutional neural fields for depth estimation from a single image

    Fayao Liu, Chunhua Shen, and Guosheng Lin. Deep con- volutional neural fields for depth estimation from a single image. CVPR, 2015

    2015

  23. [30]

    Discrete-continuous depth estimation from a single image

    Miaomiao Liu, Mathieu Salzmann, and Xuming He. Discrete-continuous depth estimation from a single image. In CVPR, 2014

    2014

  24. [31]

    U-net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. MICCAI, 2015

    2015

  25. [32]

    Monocular depth esti- mation using neural regression forest

    Anirban Roy and Sinisa Todorovic. Monocular depth esti- mation using neural regression forest. CVPR, 2016

    2016

  26. [33]

    Learn- ing depth from single monocular images

    Ashutosh Saxena, Sung H Chung, and Andrew Y Ng. Learn- ing depth from single monocular images. In NIPS, 2006

    2006

  27. [34]

    Make3d: Learning 3d scene structure from a single still image

    Ashutosh Saxena, Min Sun, and Andrew Y Ng. Make3d: Learning 3d scene structure from a single still image. PAMI, 2009

    2009

  28. [35]

    Layered depth images

    Jonathan Shade, Steven Gortler, Li-wei He, and Richard Szeliski. Layered depth images. SIGGRAPH ’98, 1998

    1998

  29. [36]

    Very deep convolutional networks for large-scale image recognition

    Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recognition. arXiv:1409.1556, 2014

    2014 arXiv

  30. [37]

    Chang, Manolis Savva, and Thomas Funkhouser

    Shuran Song, Fisher Yu, Andy Zeng, Angel X. Chang, Manolis Savva, and Thomas Funkhouser. Semantic scene completion from a single depth image. CVPR, 2017

    2017

  31. [38]

    Sudderth, and Michael J

    Deqing Sun, Erik B. Sudderth, and Michael J. Black. Lay- ered segmentation and optical flow estimation over time. CVPR, 2012

    2012

  32. [39]

    Scene parsing with object instances and occlusion ordering

    Joseph Tighe, Marc Niethammer, and Svetlana Lazebnik. Scene parsing with object instances and occlusion ordering. CVPR, 2014. 9

    2014

  33. [40]

    Efros, and Jitendra Malik

    Shubham Tulsiani, Saurabh Gupta, David Fouhey, Alexei A. Efros, and Jitendra Malik. Factoring shape, pose, and layout from the 2d image of a 3d scene. In CVPR, 2018

    2018

  34. [42]

    John Y . A. Wang and Edward H. Adelson. Representing moving images with layers.IEEE transactions on image pro- cessing, 1994

    1994

  35. [43]

    Peng Wang, Xiaohui Shen, Zhe Lin, Scott Cohen, Brian Price, and Alan L. Yuille. Towards unified depth and se- mantic prediction from a single image. CVPR, 2015

    2015

  36. [44]

    Adversarial semantic scene completion from a sin- gle depth image

    Yida Wang, David Joseph Tan, Nassir Navab, and Federico Tombari. Adversarial semantic scene completion from a sin- gle depth image. 3DV, 2018

    2018

  37. [45]

    MarrNet: 3D Shape Reconstruction via 2.5D Sketches

    Jiajun Wu, Yifan Wang, Tianfan Xue, Xingyuan Sun, William T Freeman, and Joshua B Tenenbaum. MarrNet: 3D Shape Reconstruction via 2.5D Sketches. In NIPS, 2017

    2017

  38. [46]

    Learning a Probabilistic La- tent Space of Object Shapes via 3D Generative-Adversarial Modeling

    Jiajun Wu, Chengkai Zhang, Tianfan Xue, William T Free- man, and Joshua B Tenenbaum. Learning a Probabilistic La- tent Space of Object Shapes via 3D Generative-Adversarial Modeling. In NIPS, 2016

    2016

  39. [47]

    Multi-scale continuous crfs as sequential deep networks for monocular depth estimation

    Dan Xu, Elisa Ricci, Wanli Ouyang, Xiaogang Wang, and Nicu Sebe. Multi-scale continuous crfs as sequential deep networks for monocular depth estimation. CVPR, 2017

    2017

  40. [48]

    Yi Yang, Sam Hallman, Deva Ramanan, and Charless C. Fowlkes. Layered object models for image segmentation. PAMI, 2012

    2012

  41. [49]

    Deep depth comple- tion of a single rgb-d image

    Yinda Zhang and Thomas Funkhouser. Deep depth comple- tion of a single rgb-d image. CVPR, 2018

    2018

  42. [50]

    Stereo magnification: Learning view syn- thesis using multiplane images

    Tinghui Zhou, Richard Tucker, John Flynn, Graham Fyffe, and Noah Snavely. Stereo magnification: Learning view syn- thesis using multiplane images. ACM Trans. Graph., 2018

    2018

  43. [51]

    Layoutnet: Reconstructing the 3d room layout from a single rgb image

    Chuhang Zou, Alex Colburn, Qi Shan, and Derek Hoiem. Layoutnet: Reconstructing the 3d room layout from a single rgb image. arXiv:1803.08999, 2018. 10 Supplementary Material We present additional evaluation results of our method, such as a visual comparison with [6, 41] on LDI ...

    2018 arXiv

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.