Pith. sign in

REVIEW 4 major objections 6 minor 13 references

Improving Self-Supervised Single View Depth Estimation by Masking Occlusion

T0 review · 4 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

Pith's one-line read A mask that drops occluded pixels improves self-supervised depth prediction.

desk verdict A clean, honest thesis-style paper proposing a depth-based occlusion mask for self-supervised depth, but the headline gain rests on a single training run and is close to seed noise. read the letter →

arxiv 1908.11112 v1 pith:CDTIXJQ4 submitted 2019-08-29 cs.CV

classification cs.CV
keywords self-superviseddepthestimationmonocularpredictionocclusionmaskviewsynthesisreprojectionlossKITTIbenchmarkobjectmotionimagereconstruction
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 tests whether a self-supervised single-image depth model trains better when the loss ignores only pixels that cannot be reconstructed because of occlusion, rather than choosing the best reconstruction per pixel as prior losses do. It introduces an occlusion mask built entirely from predicted depth: after projecting a target frame's pixels into a neighbouring frame, the mask compares the depth expected at each sample location with the depth actually predicted in that neighbour, and flags pixels where the observed depth is closer. Two loss functions use this mask, and on the KITTI benchmark the non-occluded minimum reprojection loss improves over the per-pixel minimum reprojection baseline on all reported metrics. The paper is honest that this does not fully confirm its original hypothesis: a non-occluded average loss did not beat the baseline, and the per-pixel minimum loss also helps by accidentally ignoring moving-object errors.

What carries the argument

The occlusion mask (Equation 7) is the load-bearing object: for each pixel projected from the target frame into an adjacent frame, it takes the depth expected at the projected location (from the target's depth prediction and the pose network) and the depth actually sampled at that location from the adjacent frame's predicted depth map. The pixel is masked out if the observed depth is closer than the expected depth scaled by (1 - tolerance), or if the projected coordinate falls outside the image. This mask carries the argument because it removes from training exactly the pixels whose reconstruction is unreliable due to occlusion, without adding learnable parameters or requiring gradients through the adjacent-frame depth prediction. It is inserted into two photometric losses, one that averages non-occluded reconstructions and one that keeps the per-pixel minimum selection but adds the mask as an error penalty.

What would settle it

Train the identical baseline with an occlusion mask computed from ground-truth depth maps of the adjacent frames instead of predicted ones; if the accuracy gain disappears or reverses, the reported improvement comes from the depth-variability comparison rather than from removing true occlusions. Alternatively, label occluded regions on a KITTI subset and measure the mask's precision and recall during the first training epochs.

Watch

Extended reading notes

Core claim

The central claim is that explicitly masking occlusion effects during training improves single-view depth prediction, and that this can be done with a parameter-free mask derived from the depth network's own predictions. The mask compares the depth expected at a reprojected sample location with the depth observed there in the adjacent frame's predicted depth map; if the observed depth is closer by more than a tolerance, the pixel is treated as occluded and removed from the photometric loss. Incorporated into a minimum-reprojection loss, the mask raises accuracy on KITTI across all seven reported metrics, with Abs Rel going from 0.114 to 0.113 and RMSE from 4.874 to 4.789. A second, unexpected finding is that the per-pixel minimum reprojection baseline already ignores some reprojection error caused by object motion, not just occlusion, which explains why simply averaging non-occluded regions did not outperform it.

Load-bearing premise

The mask assumes the depth predictions of the adjacent frames are accurate enough that a closer-than-expected sampled depth really means occlusion; early in training, when those predictions are poor, the mask can drop valid pixels or keep occluded ones, and the tolerance parameter was chosen under an accidental training-mode setting that may have raised it too high.

Editorial extensions

If this is right

  • Incorporating the occlusion mask into a per-pixel minimum reprojection loss improves all seven KITTI depth metrics relative to the baseline.
  • For average-reprojection losses, the mask improves most accuracy metrics but leaves Abs Rel unchanged and makes Sq Rel and RMSE slightly worse, so its benefit is not uniform across loss families.
  • The per-pixel minimum reprojection loss, not the mask, is what removes much of the object-motion error; a model that actually models motion might prefer the mask because it keeps motion error as a learning signal.
  • The mask needs no extra learnable parameters and no backprop through the adjacent frames' depth predictions, so the extra training cost is limited to computing the mask itself.

Reading between the lines

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

  • If the tolerance parameter were annealed from a high value early in training to near zero later, the mask might retain valid pixels during the fragile early phase and catch closer occluders once depth predictions sharpen.
  • The method could be tested on a model that explicitly handles object motion; the paper's own discussion predicts the occlusion mask should then be more helpful than per-pixel minimum reprojection, because it preserves motion-related photometric error.
  • A strong test of the mask itself is to compare masks built from predicted depths with masks built from ground-truth depths of adjacent frames; this would separate the value of occlusion masking from the accuracy of the depth predictions it relies on.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper addresses occlusion in self-supervised monocular depth estimation trained with view synthesis. It proposes an occlusion mask computed from predicted depth maps by comparing expected depth at reprojected sample locations with the sampled depth from adjacent frame depth predictions (Eq. 7). Two new photometric losses incorporate this mask: non-occluded average reprojection (Eq. 8) and non-occluded minimum reprojection (Eq. 9). Experiments on KITTI (Table 1) compare these against baselines from monodepth2. The paper claims that incorporating the mask improves performance and that per-pixel min reprojection additionally ignores object-motion errors. It also reports that its original hypothesis—that ignoring only occluded regions yields more useful training signal—is not confirmed.

Significance. If the central empirical claim were fully established, the occlusion mask would be a simple, parameter-efficient addition to self-supervised depth estimation pipelines, since it uses only predicted geometry and adds no learned parameters. The paper is also valuable for its clear visualization of what per-pixel minimum reprojection actually ignores (Figs. 7 and 8), and it openly discusses a failed hypothesis. However, the current evidence is not yet convincing: single-run training, small margins, and an accidentally chosen tolerance parameter mean the quantitative improvement could be within stochastic variation.

major comments (4)
  1. [Section 3.1.1, Table 1] The central claim (i) is supported only by a single training run per variant, and the key comparison (per-pixel min vs non-occluded min) shows a 0.001 Abs Rel and 0.085 RMSE difference. In this training regime, run-to-run variation from random seeds and checkpoint selection is typically of the same order, so without repeated runs, variance estimates, or a multiple-seed comparison, the reported improvement cannot be distinguished from noise.
  2. [Section 3.1.1, Table 1a] The average-reprojection baseline is taken directly from Table 2 of [6] rather than re-trained in the author's setup, which mixes implementations. Moreover, the non-occluded average reprojection is worse than this borrowed baseline on Sq Rel (0.936 vs 0.878) and RMSE (4.908 vs 4.846), so the statement in §3.1.1 that "Both types of loss functions (a, b) get improved result when occlusion mask is incorporated" is not supported by the table.
  3. [Appendix A, Eq. (7)] The tolerance parameter was set to 0.3 while the network was accidentally kept in training mode, which the authors state may have overestimated depth variability. No sensitivity analysis over this parameter is reported, so a reader cannot tell whether the improvements in Table 1 are robust to a different, defensible tolerance value or to a mask computed from a model in evaluation mode.
  4. [Section 2.3.1, Eq. (7)] The mask uses the same predicted depth maps that are being optimized, creating a bootstrap dependence; early in training, unreliable depth predictions may cause the mask to drop valid pixels or retain occluded ones. A concrete test would be to compare masks generated online against masks generated with a fixed, pretrained depth network or against ground-truth visibility maps, and to report how the training loss and final metrics respond.
minor comments (6)
  1. [Section 2.3.1, Eq. (7)] The text defines the sampled depth as z_{t'}^* but the equation uses z_{t->t'}^*; consistent notation should be used.
  2. [Section 2.3.3, Eq. (9)] Adding (1 - omega) to a photometric error already in [0,1] gives a loss in [0,2], which may change the effective behavior of the automask threshold in Eq. (4); this should be discussed or the normalization clarified.
  3. [Section 3.1.2, Fig. 8] Claim (ii) is based on a single randomly selected training example; a quantitative measure over many examples (e.g., fraction of pixels selected from the previous vs next frame on moving objects) would make the claim more robust.
  4. [Section 2.1.2, Eq. (1)] The projection notation uses z_t both as a homogeneous coordinate and as depth; consider clarifying the coordinate frame to avoid confusion.
  5. [Section 3.1] The manuscript never reports training hyperparameters such as image resolution, batch size, and number of epochs; these details are needed for reproducibility.
  6. [Sections 4 and Appendix A] The text frequently refers to "this thesis" and to informal details such as forgetting to switch the network to evaluation mode; while the transparency is appreciated, the phrasing should be adjusted for a journal venue.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the occlusion mask's dependence on the trained depth predictions is a bootstrap training signal, not a derivation of the evaluated result, and the KITTI evaluation is externally grounded.

full rationale

The paper's central claim—that adding the occlusion mask of Eq. (7) to the photometric loss improves KITTI depth metrics—is not derived from the mask itself. The mask is a binary weight computed by comparing the projected depth z_{t→t'} with the depth z_{t'*} sampled from the adjacent frame's predicted depth map; this is a geometric consistency check on the network's own predictions. That creates a bootstrap dependence during training, since poor depth maps can produce unreliable masks, but it does not make the reported improvement equivalent to an input by construction. The evaluation in Table 1 is against external KITTI ground truth, so the empirical claim stands or falls independently of the mask's formulation. The paper even reports that its original hypothesis was not confirmed: the non-occluded average reprojection loss fails to beat the per-pixel minimum reprojection baseline, which is the opposite of what a forced or self-validating construction would predict. The tolerance parameter (Appendix A) was hand-set, and the author candidly notes the network was accidentally left in training mode; this is a methodological weakness and a correctness risk, not a circular reduction. There are no load-bearing self-citations: the baseline [6] is an external, published implementation, and no uniqueness theorem or prior-work ansatz is imported to forbid alternatives. No equation in the paper reduces to its own input or renames a fitted parameter as a prediction. Under the stated criteria, no significant circularity is present.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The method introduces one free parameter (tolerance) and relies on the geometric assumption that observed closer depth at a reprojection sample indicates occlusion. The smoothness weight lambda and SSIM weight alpha are inherited from the baseline [6], not new.

free parameters (1)
  • tolerance = 0.3
    Threshold in Equation 7 controlling how much closer the sampled depth must be to count as occlusion. Chosen by visual inspection of masks; the author notes in Appendix A that it may be overestimated because eval mode was forgotten during selection.
assumptions (3)
  • domain assumption Image reconstruction assumes a static scene, no occlusion/disocclusion, and Lambertian surfaces.
    The self-supervised training signal is only valid under these conditions; the occlusion mask tries to repair one violation (occlusion) but not object motion or non-Lambertian surfaces. Stated in Section 2.1.2.
  • domain assumption The camera intrinsics matrix K is known and fixed.
    The reprojection equation (1) requires calibration; KITTI provides it.
  • domain assumption A sample in the adjacent frame whose predicted depth is closer than the projected expected depth indicates occlusion of the target pixel.
    This is the core geometric heuristic behind Equation 7; it presumes the adjacent-frame depth map is sufficiently accurate for the comparison to be meaningful.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Improving Self-Supervised Single View Depth Estimation by Masking Occlusion." pith.science (2026). https://pith.science/paper/CDTIXJQ4

@misc{pith2026190811112,
  author       = {Pith},
  title        = {Pith review of: Improving Self-Supervised Single View Depth Estimation by Masking Occlusion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CDTIXJQ4}},
  note         = {Machine review of arXiv:1908.11112}
}
read the original abstract

Single view depth estimation models can be trained from video footage using a self-supervised end-to-end approach with view synthesis as the supervisory signal. This is achieved with a framework that predicts depth and camera motion, with a loss based on reconstructing a target video frame from temporally adjacent frames. In this context, occlusion relates to parts of a scene that can be observed in the target frame but not in a frame used for image reconstruction. Since the image reconstruction is based on sampling from the adjacent frame, and occluded areas by definition cannot be sampled, reconstructed occluded areas corrupt to the supervisory signal. In previous work arXiv:1806.01260 occlusion is handled based on reconstruction error; at each pixel location, only the reconstruction with the lowest error is included in the loss. The current study aims to determine whether performance improvements of depth estimation models can be gained by during training only ignoring those regions that are affected by occlusion. In this work we introduce occlusion mask, a mask that during training can be used to specifically ignore regions that cannot be reconstructed due to occlusions. Occlusion mask is based entirely on predicted depth information. We introduce two novel loss formulations which incorporate the occlusion mask. The method and implementation of arXiv:1806.01260 serves as the foundation for our modifications as well as the baseline in our experiments. We demonstrate that (i) incorporating occlusion mask in the loss function improves the performance of single image depth prediction models on the KITTI benchmark. (ii) loss functions that select from reconstructions based on error are able to ignore some of the reprojection error caused by object motion.

Figures

Figures reproduced from arXiv: 1908.11112 by the authors.

Figure 1
Figure 1. Effects of occlusion on the reconstructed image. Top: The image for which the depth prediction is made. Middle: The next video frame, captured 100ms later. Bottom: Reconstruction of the first image made by sampling from the next video frame. The lane markings at the bottom of the image are incorrectly re￾constructed because they are not visible in the next frame. The tree on the right-hand side of the road appears t… view at source ↗
Figure 2
Figure 2. Top: Image for which the depth prediction is made and target for image reconstruction. Second row: Reconstructions of the above image, made from the next and previous video frames re￾spectively. Third row: Map of each reconstructions pixels that are used/ignored (white/black) by minimum reprojection loss. Bot￾tom row: Examples of occlusion mask for each reconstruction, black areas are affected by occlusion and shoul… view at source ↗
Figure 3
Figure 3. (a) Depth network A standard, fully convolutional, U￾Net is used to predict depth. (b) Pose network Pose between a pair of frames is predicted with a separate pose network. Image from [6]. The loss used for training is based on warping nearby frames, to create a new image that conforms (as best as pos￾sible) to the image for which the depth map is being pre￾dicted. This ‘warping’ means that each pixel is sampled at … view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Example of image reconstruction steps. For an image (a), a depth map prediction (b&d) together with a camera transfor￾mation prediction is used to compute the sample locations (g) in the adjacent image (e) that can be used to reconstruct (h) the target image . Effects …
Figure 5
Figure 5. Figure 5: Example of the photometric error Top: The recon￾struction target image. Middle: Reconstruction made from the next video frame. Bottom: The calculated photometric error be￾tween the target and the reconstruction. Selecting the reconstruction with lowest error at that pi…
Figure 6
Figure 6. Figure 6: Example of occlusion mask computation steps. To make a reconstruction (i) of frame (a) from an adjacent frame (b) , sample locations (d) are computed. This calculation also provides each sample locations expected depth (e). These expected depths (e) can be compared wit…
Figure 7
Figure 7. Figure 7: Visualizing photometric losses: Example with the largest difference between between the per-pixel minimum reprojection loss and the non-occluded average reprojection loss. (a) The reconstruction target. (b&c) Reconstructions made from video frames adjacent to the targe…
Figure 8
Figure 8. Figure 8: Visualizing photometric losses: Example that was randomly selected. In (n) we can see that the per-pixel minimum reprojection loss is able to ignore some of the reprojection error caused by object motion. mask is added to the loss, and consequently will be ignored by t…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

13 extracted references · 7 canonical work pages

  1. [6]

    the surfaces appear uniformly bright from all directions of view so that the photo-consistency error is meaningful 11

  2. [1]

    Casser, S

    V . Casser, S. Pirk, R. Mahjourian, and A. Angelova. Depth Prediction Without the Sensors: Leveraging Structure for Unsupervised Learning from Monocu- lar Videos. arXiv:1811.06152 [cs], Nov. 2018. arXiv: 1811.06152. 2

  3. [2]

    Eigen, C

    D. Eigen, C. Puhrsch, and R. Fergus. Depth Map Pre- diction from a Single Image using a Multi-Scale Deep Network. In Z. Ghahramani, M. Welling, C. Cortes, N. D. Lawrence, and K. Q. Weinberger, editors, Ad- vances in Neural Information Processing Systems 27 , pages 2366–2374. Curran Associates, Inc., 2014. 8

  4. [3]

    R. Garg, V . K. BG, G. Carneiro, and I. Reid. Unsuper- vised CNN for Single View Depth Estimation: Geom- etry to the Rescue. arXiv:1603.04992 [cs], Mar. 2016. arXiv: 1603.04992. 1

  5. [4]

    Geiger, P

    A. Geiger, P. Lenz, and R. Urtasun. Are we ready for autonomous driving? The KITTI vision bench- mark suite. In 2012 IEEE Conference on Computer Vi- sion and Pattern Recognition, pages 3354–3361, June

  6. [5]

    Godard, O

    C. Godard, O. Mac Aodha, and G. J. Brostow. Unsu- pervised Monocular Depth Estimation with Left-Right Consistency. arXiv:1609.03677 [cs, stat], Sept. 2016. arXiv: 1609.03677. 5

  7. [7]

    Godard, O

    C. Godard, O. Mac Aodha, M. Firman, and G. Bros- tow. Digging Into Self-Supervised Monocular Depth Estimation. arXiv:1806.01260 [cs, stat] , June 2018. arXiv: 1806.01260. 1, 2, 3, 4, 5, 6, 8

  8. [8]

    K. He, G. Gkioxari, P. Doll ´ar, and R. Girshick. Mask R-CNN. arXiv:1703.06870 [cs], Mar. 2017. arXiv: 1703.06870. 2

Show all 13 references
  1. [9]

    Jaderberg, K

    M. Jaderberg, K. Simonyan, A. Zisserman, and k. kavukcuoglu. Spatial Transformer Networks. In C. Cortes, N. D. Lawrence, D. D. Lee, M. Sugiyama, and R. Garnett, editors, Advances in Neural Informa- tion Processing Systems 28, pages 2017–2025. Curran Associates, Inc., 2015. 4

  2. [10]

    Mahjourian, M

    R. Mahjourian, M. Wicke, and A. Angelova. Un- supervised Learning of Depth and Ego-Motion from Monocular Video Using 3d Geometric Con- straints. arXiv:1802.05522 [cs] , Feb. 2018. arXiv: 1802.05522. 2, 6

  3. [11]

    C. Wang, J. M. Buenaposada, R. Zhu, and S. Lucey. Learning Depth from Monocular Videos Using Di- rect Methods. 2018 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , pages 2022– 2030, June 2018. 5

  4. [12]

    T. Zhou, M. Brown, N. Snavely, and D. G. Lowe. Un- supervised Learning of Depth and Ego-Motion from Video. arXiv:1704.07813 [cs] , Apr. 2017. arXiv: 1704.07813. 1, 2, 3, 4, 5, 8

  5. [13]

    tolerance

    Zhou Wang, A. C. Bovik, H. R. Sheikh, and E. P. Si- moncelli. Image quality assessment: from error vis- ibility to structural similarity. IEEE Transactions on Image Processing, 13(4):600–612, Apr. 2004. 5 12 Appendices A. Occlusion mask - parameter value In this work a “tolera...

Pith tools

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