Pith. sign in

REVIEW 4 major objections 4 minor 35 references

LiRCDepth: Lightweight Radar-Camera Depth Estimation via Knowledge Distillation and Uncertainty Guidance

T0 review · 4 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A lightweight radar-camera depth model with 12.65M parameters matches heavier models when knowledge distillation and uncertainty guidance are used during training.

desk verdict Useful distillation recipe for lightweight radar-camera depth, but the headline uncertainty loss is written backwards relative to its equations; needs a fix before the reported gains can be interpreted. read the letter →

arxiv 2412.16380 v2 pith:DRDUBMAW submitted 2024-12-20 cs.CV eess.IV

Huawei Sun , Nastassia Vysotskaya , Tobias Sukianto , Hao Feng , Julius Ott , Xiangyuan Peng , Lorenzo Servadei , Robert Wille This is my paper · ORCID
classification cs.CVeess.IV
keywords radar-cameradepthestimationknowledgedistillationuncertainty-rectifiedlosslightweightnetworkautonomousdrivingnuScenesMobileNetV2
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 show that a small radar-camera depth estimation network does not have to sacrifice accuracy: by copying three kinds of knowledge from a large teacher model during training, the lightweight student can nearly match the teacher while using about one fifth of the parameters and computation. The authors call the student LiRCDepth and demonstrate it on the nuScenes driving dataset. The central quantitative claim is that with knowledge distillation and an uncertainty-rectified depth loss, LiRCDepth improves mean absolute error by 6.6% and root mean square error by 3.8% over the same model trained without distillation. A sympathetic reader would care because this is the first distillation recipe applied to radar-camera depth estimation, and it points toward accurate depth perception that could run in real time on embedded hardware.

What carries the argument

The load-bearing mechanism is the uncertainty map $U = \mathbf{1}_{H \times W \times 1} - \exp\!\left(-\frac{|D_{\mathrm{pred}} - D_{\mathrm{gt}}|}{\beta\,|D_{\mathrm{pred}} + D_{\mathrm{gt}}|}\right)$ from Eq. 4, a Laplace-inspired confidence weight used twice: element-wise in the inter-depth distillation loss to emphasize pixels where the student's intermediate depths deviate most from the teacher's local-planar-guidance (LPG) depths, and after a softmax over the concatenated single-scan and dense-depth uncertainty maps in the final depth loss. Around this sit three distillation losses: pixel-wise $\ell^1$ transfer of single-modal image and radar features, an L2 loss on pairwise pixel-similarity maps of decoding features, and the uncertainty-weighted intermediate-depth loss. An affinity module aligns the student's channel dimensions to the teacher's before these losses are computed.

What would settle it

On the same validation inputs, compare the per-pixel values of the student's intermediate depth maps with the teacher's LPG depth maps; if their mean, scale, or error distribution differ strongly, the uncertainty weights in Eq. 4 are mixing mismatched quantities, and the reported gains should be attributed to teacher-specific bias rather than transferable depth knowledge.

Watch

Extended reading notes

Core claim

The paper establishes that a lightweight radar-camera depth estimator can reach near-teacher accuracy by transferring three kinds of knowledge from a heavyweight CaFNet teacher: pixel-wise single-modal features, pairwise similarity structure in decoder features, and uncertainty-weighted intermediate depth maps. The student, LiRCDepth, uses a MobileNetV2 encoder and a UNet-based decoder with 12.65M parameters and 121G FLOPs, about 80% fewer parameters than the 62.25M-parameter teacher. On nuScenes, the fully distilled student improves MAE by 6.6% and RMSE by 3.8% over direct training, matching the 80-meter row of Table I where MAE falls from 2.152 to 2.009 and RMSE from 4.801 to 4.617. The paper also claims that the proposed uncertainty-rectified depth loss, which softmax-weights errors against accumulated dense LiDAR and single-scan depth, improves over plain L1 loss and contributes to the final result.

Load-bearing premise

The load-bearing premise is that the student's intermediate depth maps are directly comparable to the teacher's LPG depth maps even though the two decoders are built differently; if their depth statistics differ, the uncertainty-weighted distillation could transfer decoder-specific artifacts rather than useful geometry.

Editorial extensions

If this is right

  • The distilled student runs at 0.069 seconds per frame with 121G FLOPs and 12.65M parameters, about 20% of the teacher's FLOPs and roughly 80% fewer parameters.
  • On the nuScenes test set, knowledge distillation improves the student's MAE by 6.6% and RMSE by 3.8% over direct training, bringing it close to the 62.25M-parameter teacher.
  • Ablations show the uncertainty weighting matters: removing uncertainty from the inter-depth distillation raises MAE from 2.009 to 2.072 and RMSE from 4.617 to 4.703 at the 80-meter range.
  • Because this is the first application of knowledge distillation to radar-camera depth estimation, the same teacher-student recipe can be carried over to other multimodal driving perception tasks.

Reading between the lines

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

  • The uncertainty map in Eq. 4 could be exported as a per-pixel confidence signal for downstream planning or sensor fusion, although the paper only uses it as a training weight.
  • The paper leaves the hyperparameter $\beta$ in Eq. 4 unspecified; a sensitivity sweep would show whether the reported gains are robust across a range of values or concentrated on a tuned point.
  • The same three-channel distillation recipe could be tried on other sparse-to-dense sensor pairs, such as camera with sparse LiDAR or camera with ultrasonic sensors.
  • Multi-teacher distillation, which the paper names as future work, is a natural next step because the three loss channels are already separated by feature level and could average teacher targets independently.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes LiRCDepth, a lightweight radar-camera depth estimation model with a MobileNetV2 backbone and a UNet-based decoder, trained with knowledge distillation from a heavier CaFNet teacher. The distillation transfers single-modal features, decoder structure, and intermediate depth maps, and the paper also introduces an Uncertainty-Rectified Depth Loss (URDL) that combines dense accumulated LiDAR depth and single-scan LiDAR depth. On the nuScenes official test set, LiRCDepth with distillation reports 12.65M parameters and 121G FLOPs, with improvements over the no-distillation baseline (e.g., MAE 2.009 vs 2.152 at 80m, a 6.6% gain), and achieves results competitive with heavier models such as CaFNet and RadarNet.

Significance. If the claims hold, the paper is a useful contribution: it is the first knowledge-distillation framework for radar-camera depth estimation, it demonstrates a large efficiency reduction (roughly 80% fewer parameters and 5x fewer FLOPs than the teacher), and it evaluates on the official nuScenes test set with a public code link. The distillation design is more comprehensive than straightforward feature mimicry, and the ablation study in Table III shows consistent gains from each distillation component. However, the central URDL formulation contains a sign inconsistency that must be resolved before the reported ablations can be interpreted as validating the proposed mechanism.

major comments (4)
  1. [II.E, Eq. (4)-(6)] The uncertainty-rectified depth loss as written implements the opposite of the described intent. Eq. (4) defines U as monotonically increasing in |Dpred - Dgt|, and Eq. (6) multiplies the L1 depth terms by U. Consequently, a pixel where the accumulated dense depth Dd is corrupted by ego-motion error and disagrees with a correct prediction receives a large Ud and hence a large loss weight, whereas Section II.E states the loss should "decrease the loss weight for the compensated pixel from other frames." Similarly, reliable single-scan pixels, where Ds is close to the prediction, receive small Us and are down-weighted, contradicting the stated focus on Ds. The softmax normalization only rescales Ud and Us and does not reverse the ordering. An implementation faithful to Eqs. (4)-(6) would up-weight noisy supervision, so the URDL ablation in Table II does not validate the mechanism described in the text.
  2. [II.D/II.E, Eq. (4) and Eq. (7)] The scale parameter beta in Eq. (4) is never reported, and the distillation weights gamma_1 through gamma_4 in Eq. (7) are not given in the implementation details. Because URDL and the four distillation losses are central to the claimed gains, the experiments as described are not reproducible without these values. The provided code link may resolve this, but the manuscript should state the hyperparameters used.
  3. [III.D, Table II] All ablations appear to be single runs with no variance or seed information. Several improvements are small (for example, AbsREL is unchanged at 0.105 in the L1 vs URDL rows, and the RMSE gain is 4.892 to 4.801), so without multiple seeds it is difficult to assess whether the uncertainty weighting provides a statistically meaningful improvement.
  4. [III.A, Table I] The teacher CaFNet was retrained using the proposed URDL, and the paper does not report the original teacher's performance or ablate whether distillation from the retrained teacher versus the original teacher affects the student. Since the student's final performance is measured against a teacher that was itself modified by the proposed loss, part of the student's gain could be inherited from the teacher's own improvement rather than from the distillation losses themselves.
minor comments (4)
  1. [Fig. 2 caption] The caption contains a typo: "Colume" should be "Column."
  2. [Table III header] The table caption reads "Table. III" with a stray period; it should be "Table III."
  3. [Fig. 1] The figure labels "MobineNetV2" should be "MobileNetV2."
  4. [Eq. (3)] The normalization factor 1/(W^i H^i)^2 is ambiguous: it is not clear whether the pairwise similarity maps are averaged over all pixel pairs or whether the denominator should instead be (W^i H^i)^2 - 1 or a different count, since the diagonal terms p=q are always 1.

Circularity Check

0 steps flagged · score 2.0 of 10

No derivation-chain circularity; the self-cited CaFNet teacher and URDL are empirical components, not outcomes of the derivation.

full rationale

The paper is an empirical systems paper. The claimed result - distillation from a 62.25M-parameter CaFNet teacher improves the 12.65M-parameter LiRCDepth student by 6.6% MAE on nuScenes - is tested against the official test set and supported by ablations (Tables I-III). Reading Eqs. 1-7, each loss is a standard training objective: single-modal feature distance (Eq. 1), pairwise structural similarity (Eq. 3), a weighted intermediate-depth imitation loss (Eq. 5), and a weighted L1 depth loss (Eq. 6). The weights in Eq. 6 are computed from the prediction and the same ground-truth maps used by the loss; this is loss reweighting, not a fitted parameter renamed as a prediction, and it does not make the outcome equal to the input by construction. The teacher is the authors' prior CaFNet and was retrained with the proposed URDL; this is a real dependency on self-cited work, but the teacher is a separate, heavier model whose outputs are externally evaluated on nuScenes, so the dependency is mild rather than load-bearing. Two caveats do not change the circularity verdict: the value of beta in Eq. 4 is not reported, and Section II.E's statement that URDL should 'decrease the loss weight for the compensated pixel from other frames' appears contradicted by Eq. 4's monotonic increase in |Dpred-Dgt| and Eq. 6's multiplication by that increase. These are correctness or implementation concerns, outside the circularity analysis.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim relies on the teacher being a reliable oracle, on the dense LiDAR accumulation being usable, on the affinity module successfully aligning different backbones, and on an unstated beta in the uncertainty formula. These are domain assumptions or ad hoc modeling choices rather than derived results.

free parameters (2)
  • beta (uncertainty scale)
    Appears in Eq. 4 as the denominator scale; no value or default is reported, yet it determines the shape of all uncertainty maps used in both the inter-depth distillation and the URDL loss.
  • gamma_1..gamma_4 (loss weights)
    In Eq. 7, the total loss is a weighted sum with gamma factors, but the text does not give their values or search range.
assumptions (4)
  • domain assumption The teacher CaFNet (retrained with URDL) provides reliable feature and depth targets for distillation.
    Section II.A uses CaFNet as the teacher; Section III.A states it was retrained with the proposed URDL, so the student's supervision depends on a teacher that has already absorbed the proposed loss.
  • domain assumption Dense LiDAR depth Dd generated by accumulating neighboring point clouds with ego-motion is a usable supervision signal despite introduced errors.
    Section III.A adopts the accumulation strategy from [12]; Section II.E acknowledges these errors and adds Ds to mitigate them.
  • ad hoc to paper A 1x1 convolution plus ReLU (affinity module) can align MobileNetV2 feature channels to ResNet34 teacher features across scales.
    Section II.A introduces the affinity module to match channel sizes; if the feature statistics are too different, the L1 distillation losses in Eq. 1 would be meaningless.
  • ad hoc to paper The Laplace-inspired uncertainty weighting in Eq. 4 with hyperparameter beta yields a useful per-pixel loss weight.
    Section II.D defines U with beta unspecified; the behavior of the weighting is not justified beyond analogy to a Laplace probability density function.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LiRCDepth: Lightweight Radar-Camera Depth Estimation via Knowledge Distillation and Uncertainty Guidance." pith.science (2026). https://pith.science/paper/DRDUBMAW

@misc{pith2026241216380,
  author       = {Pith},
  title        = {Pith review of: LiRCDepth: Lightweight Radar-Camera Depth Estimation via Knowledge Distillation and Uncertainty Guidance},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DRDUBMAW}},
  note         = {Machine review of arXiv:2412.16380}
}
read the original abstract

Recently, radar-camera fusion algorithms have gained significant attention as radar sensors provide geometric information that complements the limitations of cameras. However, most existing radar-camera depth estimation algorithms focus solely on improving performance, often neglecting computational efficiency. To address this gap, we propose LiRCDepth, a lightweight radar-camera depth estimation model. We incorporate knowledge distillation to enhance the training process, transferring critical information from a complex teacher model to our lightweight student model in three key domains. Firstly, low-level and high-level features are transferred by incorporating pixel-wise and pair-wise distillation. Additionally, we introduce an uncertainty-aware inter-depth distillation loss to refine intermediate depth maps during decoding. Leveraging our proposed knowledge distillation scheme, the lightweight model achieves a 6.6% improvement in MAE on the nuScenes dataset compared to the model trained without distillation. Code: https://github.com/harborsarah/LiRCDepth

Figures

Figures reproduced from arXiv: 2412.16380 by the authors.

Figure 1
Figure 1. Model Architecture. framework. To ensure satisfactory performance, we distill three key components from the teacher to the student model, resulting in a 6.6% performance improvement compared to direct training. • Our model is evaluated on nuScenes [14], achieving results comparable to other heavy-weight algorithms. II. METHODOLOGY This section first introduces the model architecture. Then, we describe the knowledge … view at source ↗
Figure 2
Figure 2. Qualitative comparison at 80 meters depth range. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

35 extracted references · 14 canonical work pages

  1. [1]

    From big to small: Multi-scale local planar guidance for monocular depth estimation,

    J. H. Lee, M.-K. Han, D. W. Ko, and I. H. Suh, “From big to small: Multi-scale local planar guidance for monocular depth estimation,”arXiv preprint arXiv:1907.10326, 2019

    arXiv 1907

  2. [2]

    Binsformer: Revisiting adaptive bins for monocular depth estimation,

    Z. Li, X. Wang, X. Liu, and J. Jiang, “Binsformer: Revisiting adaptive bins for monocular depth estimation,” arXiv preprint arXiv:2204.00987, 2022

    arXiv 2022

  3. [3]

    P3depth: Monocular depth estimation with a piecewise planarity prior,

    V . Patil, C. Sakaridis, A. Liniger, and L. Van Gool, “P3depth: Monocular depth estimation with a piecewise planarity prior,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 1610–1621

    2022

  4. [4]

    Va-depthnet: A variational approach to single image depth prediction,

    C. Liu, S. Kumar, S. Gu, R. Timofte, and L. Van Gool, “Va-depthnet: A variational approach to single image depth prediction,” arXiv preprint arXiv:2302.06556, 2023

    arXiv 2023

  5. [5]

    Deep ordinal regression network for monocular depth estimation,

    H. Fu, M. Gong, C. Wang, K. Batmanghelich, and D. Tao, “Deep ordinal regression network for monocular depth estimation,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2018, pp. 2002–2011

    2018

  6. [6]

    Monocular depth estimation using laplacian pyramid-based depth residuals,

    M. Song, S. Lim, and W. Kim, “Monocular depth estimation using laplacian pyramid-based depth residuals,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 31, no. 11, pp. 4381–4393, 2021

    work page 2021

  7. [7]

    Learning depth with convolutional spatial propagation network,

    X. Cheng, P. Wang, and R. Yang, “Learning depth with convolutional spatial propagation network,” IEEE transactions on pattern analysis and machine intelligence, vol. 42, no. 10, pp. 2361–2379, 2019

    work page 2019

  8. [8]

    Depth estimation from monocular images and sparse radar data,

    J.-T. Lin, D. Dai, and L. Van Gool, “Depth estimation from monocular images and sparse radar data,” in 2020 IEEE/RSJ International Con- ference on Intelligent Robots and Systems (IROS) . IEEE, 2020, pp. 10 233–10 240

    2020

Show all 35 references
  1. [9]

    Radar-camera pixel depth association for depth completion,

    Y . Long, D. Morris, X. Liu, M. Castro, P. Chakravarty, and P. Narayanan, “Radar-camera pixel depth association for depth completion,” in Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 12 507–12 516

    2021

  2. [10]

    Depth estimation from monocular images and sparse radar using deep ordinal regression network,

    C.-C. Lo and P. Vandewalle, “Depth estimation from monocular images and sparse radar using deep ordinal regression network,” in 2021 IEEE International Conference on Image Processing (ICIP). IEEE, 2021, pp. 3343–3347

    2021

  3. [11]

    Depth estimation from camera image and mmwave radar point cloud,

    A. D. Singh, Y . Ba, A. Sarker, H. Zhang, A. Kadambi, S. Soatto, M. Srivastava, and A. Wong, “Depth estimation from camera image and mmwave radar point cloud,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 9275–9285

    2023

  4. [12]

    Cafnet: A confidence-driven framework for radar camera depth estimation,

    H. Sun, H. Feng, J. Ott, L. Servadei, and R. Wille, “Cafnet: A confidence-driven framework for radar camera depth estimation,” arXiv preprint arXiv:2407.00697, 2024

    2024 arXiv

  5. [13]

    Sparse beats dense: Re- thinking supervision in radar-camera depth completion,

    H. Li, M. Jing, J. Liang, H. Fan, and R. Ji, “Sparse beats dense: Re- thinking supervision in radar-camera depth completion,” arXiv preprint arXiv:2312.00844, 2023

    2023 arXiv

  6. [14]

    nuscenes: A multimodal dataset for autonomous driving,

    H. Caesar, V . Bankiti, A. H. Lang, S. V ora, V . E. Liong, Q. Xu, A. Krishnan, Y . Pan, G. Baldan, and O. Beijbom, “nuscenes: A multimodal dataset for autonomous driving,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 11 621–11 631

    2020

  7. [15]

    Multi-task cross-modality attention-fusion for 2d object detection,

    H. Sun, H. Feng, G. Stettinger, L. Servadei, and R. Wille, “Multi-task cross-modality attention-fusion for 2d object detection,” in 2023 IEEE 26th International Conference on Intelligent Transportation Systems (ITSC), 2023, pp. 3619–3626

    2023

  8. [16]

    Enhanced radar perception via multi-task learning: Towards refined data for sensor fusion applications,

    H. Sun, H. Feng, G. Mauro, J. Ott, G. Stettinger, L. Servadei, and R. Wille, “Enhanced radar perception via multi-task learning: Towards refined data for sensor fusion applications,” in 2024 IEEE Intelligent Vehicles Symposium (IV), 2024, pp. 3179–3184

    2024

  9. [17]

    Get-up: Geometric-aware depth estimation with radar points upsampling,

    H. Sun, Z. Wang, H. Feng, J. Ott, L. Servadei, and R. Wille, “Get-up: Geometric-aware depth estimation with radar points upsampling,” 2024. [Online]. Available: https://arxiv.org/abs/2409.02720

    2024 arXiv

  10. [18]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2016, pp. 770–778

    2016

  11. [19]

    Mobilenetv2: Inverted residuals and linear bottlenecks,

    M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, “Mobilenetv2: Inverted residuals and linear bottlenecks,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2018, pp. 4510–4520

    2018

  12. [20]

    Distilling the knowledge in a neural network,

    G. Hinton, O. Vinyals, and J. Dean, “Distilling the knowledge in a neural network,” arXiv preprint arXiv:1503.02531 , 2015

    2015 arXiv

  13. [21]

    Deep mutual learning,

    Y . Zhang, T. Xiang, T. M. Hospedales, and H. Lu, “Deep mutual learning,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2018, pp. 4320–4328

    2018

  14. [22]

    Fitnets: Hints for thin deep nets,

    A. Romero, N. Ballas, S. E. Kahou, A. Chassang, C. Gatta, and Y . Ben- gio, “Fitnets: Hints for thin deep nets,” arXiv preprint arXiv:1412.6550, 2014

    2014 arXiv

  15. [23]

    Paying more attention to attention: Improving the performance of convolutional neural networks via atten- tion transfer,

    S. Zagoruyko and N. Komodakis, “Paying more attention to attention: Improving the performance of convolutional neural networks via atten- tion transfer,” arXiv preprint arXiv:1612.03928 , 2016

    2016 arXiv

  16. [24]

    Mimicking very efficient network for object detection,

    Q. Li, S. Jin, and J. Yan, “Mimicking very efficient network for object detection,” in Proceedings of the ieee conference on computer vision and pattern recognition , 2017, pp. 6356–6364

    2017

  17. [25]

    Crkd: Enhanced camera-radar object detection with cross-modality knowledge distillation,

    L. Zhao, J. Song, and K. A. Skinner, “Crkd: Enhanced camera-radar object detection with cross-modality knowledge distillation,” in Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 15 470–15 480

    2024

  18. [26]

    Learning efficient object detection models with knowledge distillation,

    G. Chen, W. Choi, X. Yu, T. Han, and M. Chandraker, “Learning efficient object detection models with knowledge distillation,” Advances in neural information processing systems , vol. 30, 2017

    2017

  19. [27]

    Knowledge distillation for semantic segmenta- tion using channel and spatial correlations and adaptive cross entropy,

    S. Park and Y . S. Heo, “Knowledge distillation for semantic segmenta- tion using channel and spatial correlations and adaptive cross entropy,” Sensors, vol. 20, no. 16, p. 4616, 2020

    2020

  20. [28]

    Structured knowledge distillation for semantic segmentation,

    Y . Liu, K. Chen, C. Liu, Z. Qin, Z. Luo, and J. Wang, “Structured knowledge distillation for semantic segmentation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2019, pp. 2604–2613

    2019

  21. [29]

    Knowledge adaptation for efficient semantic segmentation,

    T. He, C. Shen, Z. Tian, D. Gong, C. Sun, and Y . Yan, “Knowledge adaptation for efficient semantic segmentation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 578–587

    2019

  22. [30]

    Boost- ing monocular depth estimation with lightweight 3d point fusion,

    L. Huynh, P. Nguyen, J. Matas, E. Rahtu, and J. Heikkil ¨a, “Boost- ing monocular depth estimation with lightweight 3d point fusion,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 12 767–12 776

    2021

  23. [31]

    Knowledge distillation for fast and accurate monocular depth estimation on mobile devices,

    Y . Wang, X. Li, M. Shi, K. Xian, and Z. Cao, “Knowledge distillation for fast and accurate monocular depth estimation on mobile devices,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 2457–2465

    2021

  24. [32]

    Lightweight depth completion network with local similarity-preserving knowledge distillation,

    Y . Jeong, J. Park, D. Cho, Y . Hwang, S. B. Choi, and I. S. Kweon, “Lightweight depth completion network with local similarity-preserving knowledge distillation,” Sensors, vol. 22, no. 19, p. 7388, 2022

    2022

  25. [33]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” in Medical Image Computing and Computer-Assisted Intervention–MICCAI 2015: 18th International Conference, Munich, Germany, October 5-9, 2015, Proceedings, Part III

    2015

  26. [34]

    Springer, 2015, pp. 234–241

    2015

  27. [35]

    Pytorch: An imperative style, high- performance deep learning library,

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga, A. Desmaison, A. K ¨opf, E. Yang, Z. DeVito, M. Raison, A. Tejani, S. Chilamkurthy, B. Steiner, L. Fang, J. Bai, and S. Chintala, “Pytorch: An imperative style, high-...

    1912 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.