Pith. sign in

REVIEW 4 major objections 6 minor 63 references

Fully Exploiting Vision Foundation Model's Profound Prior Knowledge for Generalizable RGB-Depth Driving Scene Parsing

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

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

Pith's one-line read This paper claims that a frozen vision foundation model plus a lightweight depth side adapter can outperform all compared RGB-D scene parsers, reaching 84.74% mIoU on Cityscapes.

desk verdict A practical adapter paper whose real contribution is showing VFM-predicted relative depth can replace metric depth in RGB-D parsing; the SOTA claims need tighter baseline control. read the letter →

arxiv 2502.06219 v1 pith:FWQJZ4EB submitted 2025-02-10 cs.CV

Sicen Guo , Tianyou Wen , Chuang-Wei Liu , Qijun Chen , Rui Fan This is my paper · ORCID
classification cs.CV
keywords RGB-Dsceneparsingvisionfoundationmodelssideadapterrelativedepthestimationheterogeneousfeaturefusiondrivingsemanticsegmentationtransformer
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 argues that a vision foundation model (VFM) trained only on RGB images can be turned into a strong RGB-depth driving-scene parser without retraining the foundation model. The proposed network, HFIT, freezes the VFM backbone and adds a lightweight side adapter that digests the RGB image together with a relative depth map produced by the VFM itself. On Cityscapes and KITTI Semantics, HFIT reports the best mIoU among all compared single-modal and data-fusion methods, reaching 84.74% on Cityscapes with a Depth Anything V1 backbone. The paper's key practical claim is that relative depth, which any modern VFM can supply, is nearly as useful as metric depth: swapping in relative depth costs less than 0.12% mIoU. If the claim holds, depth sensors become optional for this task, because the adapter needs no LiDAR-accurate depth.

What carries the argument

The load-bearing object is the HFIT side adapter, made of three modules: the Duplex Spatial Prior Extractor (DSPE), which uses two identical EfficientNet blocks to build a multi-scale pyramid from RGB and relative depth; the Recalibrated Heterogeneous Feature Fusion (RHFF) module, which computes per-pixel confidence maps from both branches and gates the depth priors by multiplying with the complement of the ViT confidence before injecting them into the ViT with cross-attention; and the Holistic Gated Feature Integration (HGFI) module, which reweights multi-level features from both branches so lower-level detail flows upward only when gates open. The relative-depth input is itself part of the machinery: because the VFM generates the depth map, the system never needs calibrated metric depth.

What would settle it

Retrain the strongest ViT-Adapter baseline under HFIT's exact protocol (same iterations, crop size, decoder, and depth supervision) and check whether HFIT's reported 2.6-point mIoU lead on Cityscapes and 0.96-point lead on KITTI survive.

Watch

Extended reading notes

Core claim

The central discovery is that the profound prior knowledge of a frozen vision foundation model can be harvested for RGB-D driving scene parsing by treating the VFM as a fixed feature extractor and attaching a purpose-built side adapter, rather than re-training the ViT or training a second encoder. HFIT feeds RGB and the VFM's own relative depth output through a duplex EfficientNet backbone, then recalibrates those heterogeneous spatial priors with confidence maps before injecting them into the frozen transformer via cross-attention. Multi-level features are then combined by gates that move fine-grained details upward only when they are useful. The result is that HFIT outperforms all compared traditional RGB-D fusion networks, plain pre-trained VFMs, and ViT adapters, reaching 84.74% mean IoU on Cityscapes and 80.24% on KITTI; its ablation shows the gains disappear if either the recalibration weights or the gated integration are removed.

Load-bearing premise

The comparison assumes every baseline was trained under the same schedule and resolution as HFIT (20,000 iterations, 448x448 crops), but those settings are only stated for HFIT itself.

Editorial extensions

If this is right

  • Freezing the ViT and training only the adapter keeps trainable parameters low, so the approach can ride on future, larger foundation models without a matching growth in training cost.
  • Because relative depth from the VFM substitutes for metric depth within 0.12% mIoU, deployment no longer requires LiDAR or calibrated stereo on the vehicle.
  • The confidence-based gating in RHFF suggests the model can suppress unreliable depth regions, which should make the system more robust to calibration errors and noisy depth estimates.
  • On KITTI, the gains over VFM baselines are smaller than on Cityscapes, so the method's advantage grows with dataset diversity and scale.

Reading between the lines

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

  • A natural extension is to test whether the same frozen-backbone/adapter recipe transfers to other RGB-X tasks, such as RGB-thermal parsing, where the DSPE's duplex branches would ingest a different second modality.
  • The under-0.12% gap between relative and metric depth is measured only with RAFT-Stereo as the metric source; comparing against LiDAR-grade metric depth from KITTI would sharpen the claim about relative depth being sufficient.
  • One could put the confidence maps to work at inference time by thresholding low-confidence pixels and querying them for active learning, a use the paper does not discuss.
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 / 6 minor

Summary. The paper proposes HFIT, a side-adapter architecture for RGB-depth driving scene parsing that keeps a frozen vision foundation model (ViT) backbone and adds a trainable adapter composed of a Duplex Spatial Prior Extractor (DSPE), Recalibrated Heterogeneous Feature Fusion (RHFF) modules, and Holistic Gated Feature Integration (HGFI) modules. Relative depth maps predicted by a VFM (e.g., Depth Anything V2) are used as the depth input, avoiding reliance on metric depth sensors. Experiments on Cityscapes and KITTI Semantics report state-of-the-art mIoU values (84.74% on Cityscapes, 80.24% on KITTI) and ablations showing that RGB plus relative depth outperforms RGB alone and is comparable to using metric depth. The authors claim that this is the first work to adapt VFMs for RGB-D driving scene parsing and that the approach is generalizable.

Significance. If the experimental claims hold, the paper makes a useful contribution: it demonstrates a parameter-efficient way to adapt frozen vision foundation models to RGB-D scene parsing and shows that VFM-predicted relative depth can serve as a substitute for metric depth. The public release of code is a concrete strength, as is the internal protocol-matched ablation in Table III that isolates the contribution of the depth modality. However, the significance is limited by two gaps: the comparison with prior methods is not protocol-matched, so the magnitude of the reported gains is uncertain, and the 'generalizable' claim is not tested across datasets or domains. The paper also uses a heuristic confidence formulation that is not well justified and may not behave as described.

major comments (4)
  1. [IV-A, Tables I and II] The training and evaluation protocols for the baseline methods are not specified. Section IV-A reports 20,000 iterations and 448x448 random crops only for HFIT; no equivalent settings are given for OCRNet, KNet, EMANet, SNE-RoadSeg, OFFNet, MFNet, FuseNet, ViT-Adapter, or ViT-CoMer. Because HFIT's advantage over ViT-Adapter is only 2.63 mIoU on Cityscapes and 0.96 mIoU on KITTI, differences in training budget, crop size, backbone, optimizer, or evaluation protocol could erase the claimed gain. Please provide protocol-matched comparisons (ideally all methods trained under the same pipeline) or clearly state that the numbers are quoted from the original papers and cite them.
  2. [Title, Abstract, Section I, Section IV-B] The claim that HFIT is 'generalizable' is not supported by the experiments. Only Cityscapes and KITTI Semantics are used, and there is no cross-dataset or cross-domain evaluation (e.g., training on Cityscapes and testing on KITTI, or evaluating on unseen driving scenes). The authors themselves note in Section IV-B that KITTI's smaller size limits generalization. A cross-dataset experiment is needed to justify the title and abstract claims.
  3. [III-C, Eqs. (2) and (5)] The confidence maps C^V and C^S are defined as sigma(Conv(-L * log L)), where L is the output of a ReLU. The text states that 'a higher confidence level suggests the pixel significantly contributes to accurate segmentation' (Section III-C). However, for L > 1, -L log L is negative and the sigmoid output approaches 0, meaning that strong activations are treated as low-confidence. This contradicts the described semantics. Moreover, no justification or reference is given for this entropy-like transform. Since these maps directly implement the recalibration in Eq. (7), the RHFF module requires either a corrected formula, a normalization that bounds L to [0,1], or an ablation demonstrating that the particular nonlinearity is essential to the reported performance.
  4. [IV-A, III-A, Table IV] The main experimental configuration is not fully specified. Section IV-A does not state which EfficientNet variant is used in the DSPE for the main results; Table IV shows EfficientNet-B7 achieving the highest mIoU in the ablation, but it is not confirmed as the default choice. In addition, the number of interaction stages N (set to 4 in Section III-A) is not ablated, and hyperparameters such as learning rate, batch size, and optimizer are omitted from the main text. These details are necessary for reproducibility and for fair comparison with the baselines.
minor comments (6)
  1. [Tables I and II] The 'error range (%)' columns are not defined in Section IV-A; please describe how the range is computed from the three independent runs.
  2. [Section IV-B] The phrase 'increase in mIoU by 17.45-45.54%' should say 'percentage points' rather than '%' to avoid confusion.
  3. [Tables I and II, Section I] The paper argues that the adapter approach has a 'significantly lower number of trainable parameters', but no parameter counts or FLOPs are reported for HFIT or the baselines. Please include these numbers.
  4. [Table III] The rows 'Depth' and 'Relative Depth' should be explicitly defined (e.g., metric depth from RAFT-Stereo versus relative depth from Depth Anything V2), and the source of the depth maps used in the 'Depth' row should be stated.
  5. [Section IV-A] The paper states 'More details are given in the supplementary material' but does not list the learning rate, batch size, number of classes, or the train/val split. Please provide these details in the main text or ensure the supplementary material is publicly accessible.
  6. [III-C, Eqs. (2) and (5)] Please define the base of the logarithm and specify the kernel size k and dilation rate r used in the atrous convolution, as these are not stated in the text.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the empirical comparison and ablations are self-contained, and the relative-depth input is a fixed external model output, not a fitted parameter renamed as a prediction.

full rationale

This is an empirical systems paper rather than a derivation from first principles. The central claim that HFIT improves RGB-D driving scene parsing over single-modal VFMs and data-fusion baselines is supported by held-out mIoU comparisons on Cityscapes and KITTI (Tables I-II) and by internal ablations (Tables III-VI). No equation in Sections III or IV defines HFIT's output in terms of the reported evaluation metric, and no parameter is fitted to the test labels and then presented as a prediction. The relative depth maps used as inputs are generated by the fixed, externally pretrained Depth Anything V2 model; although the same model family is also used as the frozen backbone, this is a design choice rather than a circular reduction because the depth maps are not optimized against the segmentation objective and the RGB-only ablations use the identical backbone. Self-citations, such as SNE-RoadSeg [10], appear only as related-work baselines or prior system references and do not carry the load-bearing argument. Concerns about unequal training protocols for external baseline numbers are experimental-validity risks, not circularity. Therefore, no specific circular step can be exhibited, and the circularity score is 0.

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

The paper does not derive physical constants or laws; it proposes a learned network. Its central claim, empirical superiority, rests on standard deep-learning assumptions, the transferability of pretrained features, the validity of relative depth as a depth proxy, and the fairness of the comparison protocol. The last of these is the least supported. No new physical entities are introduced; the DSPE, RHFF, and HGFI modules are architectural components with learned parameters, not new particles or forces.

free parameters (4)
  • Number of interaction stages N = 4
    Chosen by hand; no ablation is provided for this hyperparameter (Section III-A).
  • Training iterations = 20000
    Used for HFIT; unclear whether baselines used the same (Section IV-A).
  • Random crop size = 448x448
    Standard choice; may not match original baseline training resolutions (Section IV-A).
  • EfficientNet variant in DSPE = EfficientNet-B7 (best)
    Selected via ablation based on mIoU (Table IV).
assumptions (5)
  • domain assumption Pretrained vision foundation models encode useful semantic and geometric priors that transfer to driving scene parsing.
    The entire method relies on this transfer; it is the stated motivation in Section I.
  • domain assumption Relative depth maps predicted by Depth Anything V2 are a sufficient substitute for metric depth in RGB-D fusion for segmentation.
    This is the paper's central thesis, tested only indirectly via the 0.12% mIoU gap in Table III.
  • ad hoc to paper The compared baseline results are reproducible and comparable under the same evaluation protocol.
    No evidence is provided that baselines were trained with identical settings; this is assumed to justify the SOTA claim (Section IV).
  • standard math Standard deep learning operations (LayerNorm, attention, deformable convolution) behave as intended in the proposed modules.
    These are established building blocks; no proof is required, but their composition is not justified other than empirically.
  • domain assumption The Cityscapes and KITTI datasets provide sufficient diversity to assess generalization.
    The paper claims generality but only evaluates on these two datasets, and even notes overfitting on the smaller KITTI set (Section IV-B).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fully Exploiting Vision Foundation Model's Profound Prior Knowledge for Generalizable RGB-Depth Driving Scene Parsing." pith.science (2026). https://pith.science/paper/FWQJZ4EB

@misc{pith2026250206219,
  author       = {Pith},
  title        = {Pith review of: Fully Exploiting Vision Foundation Model's Profound Prior Knowledge for Generalizable RGB-Depth Driving Scene Parsing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FWQJZ4EB}},
  note         = {Machine review of arXiv:2502.06219}
}
read the original abstract

Recent vision foundation models (VFMs), typically based on Vision Transformer (ViT), have significantly advanced numerous computer vision tasks. Despite their success in tasks focused solely on RGB images, the potential of VFMs in RGB-depth driving scene parsing remains largely under-explored. In this article, we take one step toward this emerging research area by investigating a feasible technique to fully exploit VFMs for generalizable RGB-depth driving scene parsing. Specifically, we explore the inherent characteristics of RGB and depth data, thereby presenting a Heterogeneous Feature Integration Transformer (HFIT). This network enables the efficient extraction and integration of comprehensive heterogeneous features without re-training ViTs. Relative depth prediction results from VFMs, used as inputs to the HFIT side adapter, overcome the limitations of the dependence on depth maps. Our proposed HFIT demonstrates superior performance compared to all other traditional single-modal and data-fusion scene parsing networks, pre-trained VFMs, and ViT adapters on the Cityscapes and KITTI Semantics datasets. We believe this novel strategy paves the way for future innovations in VFM-based data-fusion techniques for driving scene parsing. Our source code is publicly available at https://mias.group/HFIT.

Figures

Figures reproduced from arXiv: 2502.06219 by the authors.

Figure 1
Figure 1. An overview of our proposed HFIT. Through the interaction between recalibrated heterogeneous feature fusion (RHFF) modules and holistic gated feature integration (HGFI) modules, heterogeneous features are fully integrated with the profound prior features. scene parsing [10], [11], [12], [13]. However, a significant lim￾itation of these fusion models is their reliance on depth maps, which restricts their applicabilit… view at source ↗
Figure 2
Figure 2. An illustration of our proposed HFIT, consisting of (1) a plain ViT, (2) a duplex spatial prior extractor, (3) recalibrated heterogeneous feature fusion modules, and (4) holistic gated feature integration modules. C. Recalibrated Heterogeneous Feature Fusion Module The primary challenge in feature extraction for RGB-D scene parsing lies in the efficient fusion and integration of diverse heterogeneous features obtain… view at source ↗
Figure 3
Figure 3. Qualitative comparisons with SoTA scene parsing approaches on the KITTI Semantics [ et DINOV2 Depth Anything V1 Depth Anything V2 ViTAdapter [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Qualitative comparisons with SoTA scene parsing approaches on the Cityscapes [ [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Probability maps for different classes generated by VFMs and HFIT, where blue indicates high prediction confidence [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 55 canonical work pages

  1. [1]

    Segment Anything,

    A. Kirillov et al., “Segment Anything,” in Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), October 2023, pp. 4015–4026

    work page 2023

  2. [2]

    DINOv2: Learning robust visual features without supervision,

    M. Oquab et al. , “DINOv2: Learning robust visual features without supervision,” Computing Research Repository (CoRR) , vol. abs/2304.07193, 2023. [Online]. Available: https://arxiv.org/abs/2304. 07193

    arXiv 2023

  3. [3]

    Depth Anything: Unleashing the power of large-scale unlabeled data,

    L. Yang et al. , “Depth Anything: Unleashing the power of large-scale unlabeled data,” in 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 10 371–10 381

    work page 2024

  4. [4]

    ViPOcc: leveraging visual priors from vision foundation models for single-view 3d occupancy prediction,

    Y . Feng et al., “ViPOcc: leveraging visual priors from vision foundation models for single-view 3d occupancy prediction,” in Proceedings of the AAAI conference on Artificial Intelligence (AAAI) , 2025, in press

    work page 2025

  5. [5]

    Few-shot object detection with foundation models,

    G. Han and S.-N. Lim, “Few-shot object detection with foundation models,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2024, pp. 28 608–28 618

    work page 2024

  6. [6]

    Playing to vision foundation model’s strengths in stereo matching,

    C.-W. Liu et al. , “Playing to vision foundation model’s strengths in stereo matching,” IEEE Transactions on Intelligent Vehicles, 2024, DOI: 10.1109/TIV .2024.3467287. 9

    work page arXiv 2024

  7. [7]

    Joint depth prediction and semantic segmentation with multi-view SAM,

    M. Shvets et al., “Joint depth prediction and semantic segmentation with multi-view SAM,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) , 2024, pp. 1328–1338

    work page 2024

  8. [8]

    Bridging the domain gap: Self-supervised 3D scene understanding with foundation models,

    Z. Chen et al. , “Bridging the domain gap: Self-supervised 3D scene understanding with foundation models,” Advances in Neural Information Processing Systems (NeurIPS) , vol. 36, 2024

    work page 2024

Show all 63 references
  1. [9]

    HAPNet: Toward superior RGB-Thermal scene parsing via hybrid, asymmetric, and progressive heterogeneous feature fusion,

    J. Li et al. , “HAPNet: Toward superior RGB-Thermal scene parsing via hybrid, asymmetric, and progressive heterogeneous feature fusion,” IEEE Transactions on Intelligent Transportation Systems, 2024, in press

    2024

  2. [10]

    SNE-RoadSeg: Incorporating surface normal informa- tion into semantic segmentation for accurate freespace detection,

    R. Fan et al. , “SNE-RoadSeg: Incorporating surface normal informa- tion into semantic segmentation for accurate freespace detection,” in Proceedings of the European Conference on Computer Vision (ECCV) . Springer, 2020, pp. 340–356

    2020

  3. [11]

    ORFD: A dataset and benchmark for OFF-Road freespace detection,

    C. Min et al. , “ORFD: A dataset and benchmark for OFF-Road freespace detection,” in 2022 International Conference on Robotics and Automation (ICRA). IEEE, 2022, pp. 2532–2538

    2022

  4. [12]

    MFNet: Towards real-time semantic segmentation for autonomous vehicles with multi-spectral scenes,

    Q. Ha et al. , “MFNet: Towards real-time semantic segmentation for autonomous vehicles with multi-spectral scenes,” in 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . IEEE, 2017, pp. 5108–5115

    2017

  5. [13]

    FuseNet: Incorporating depth into semantic segmen- tation via fusion-based cnn architecture,

    Hazirbas et al. , “FuseNet: Incorporating depth into semantic segmen- tation via fusion-based cnn architecture,” in Proceedings of the Asian Conference on Computer Vision (ACCV). Springer, 2017, pp. 213–228

    2017

  6. [14]

    Pothole detection based on disparity transformation and road surface modeling,

    R. Fan et al. , “Pothole detection based on disparity transformation and road surface modeling,” IEEE Transactions on Image Processing , vol. 29, pp. 897–908, 2020

    2020

  7. [15]

    Graph attention layer evolves semantic segmentation for road pothole detection: A benchmark and algorithms,

    R. Fan et al., “Graph attention layer evolves semantic segmentation for road pothole detection: A benchmark and algorithms,” IEEE Transac- tions on Image Processing , vol. 30, pp. 8144–8154, 2021

    2021

  8. [16]

    These maps are made by propagation: Adapting deep stereo networks to road scenarios with decisive disparity diffusion,

    C.-W. Liu et al., “These maps are made by propagation: Adapting deep stereo networks to road scenarios with decisive disparity diffusion,” IEEE Transactions on Image Processing , 2025, in press

    2025

  9. [17]

    S 3M-Net: Joint learning of semantic segmentation and stereo matching for autonomous driving,

    Z. Wu et al. , “S 3M-Net: Joint learning of semantic segmentation and stereo matching for autonomous driving,” IEEE Transactions on Intelligent Vehicles, vol. 9, no. 2, pp. 3940–3951, 2024

    2024

  10. [18]

    Online, target-free LiDAR-camera extrinsic calibration via cross-modal mask matching,

    Z. Huang et al., “Online, target-free LiDAR-camera extrinsic calibration via cross-modal mask matching,” IEEE Transactions on Intelligent Vehicles, 2024, DOI: 10.1109/TIV .2024.3456299

    2024

  11. [19]

    Depth Anything V2,

    Yang, Lihe et al. , “Depth Anything V2,” Computing Research Repository (CoRR) , vol. abs/2406.09414, 2024. [Online]. Available: https://arxiv.org/abs/2406.09414

    2024 arXiv

  12. [20]

    Three ways to improve semantic segmentation with self-supervised depth estimation,

    L. Hoyer et al. , “Three ways to improve semantic segmentation with self-supervised depth estimation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021, pp. 11 130–11 140

    2021

  13. [21]

    Learning to relate depth and semantics for unsupervised domain adaptation,

    S. Saha et al., “Learning to relate depth and semantics for unsupervised domain adaptation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2021, pp. 8197– 8207

    2021

  14. [22]

    Efficient RGB-D semantic segmentation for indoor scene analysis,

    D. Seichter et al., “Efficient RGB-D semantic segmentation for indoor scene analysis,” in 2021 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2021, pp. 13 525–13 531

    2021

  15. [23]

    Learning common and specific features for RGB-D semantic segmentation with deconvolutional networks,

    J. Wang et al. , “Learning common and specific features for RGB-D semantic segmentation with deconvolutional networks,” in Proceedings of the European Conference on Computer Vision (ECCV) . Springer, 2016, pp. 664–679

    2016

  16. [24]

    Vision Transformer adapter for dense predictions,

    Z. Chen et al. , “Vision Transformer adapter for dense predictions,” in The Eleventh International Conference on Learning Representations (ICLR), 2023

    2023

  17. [25]

    Conv-Adapter: Exploring parameter efficient transfer learning for convnets,

    H. Chen et al. , “Conv-Adapter: Exploring parameter efficient transfer learning for convnets,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2024, pp. 1551– 1561

    2024

  18. [26]

    CANet: Co-attention network for RGB-D semantic segmentation,

    H. Zhou et al. , “CANet: Co-attention network for RGB-D semantic segmentation,” Pattern Recognition, vol. 124, p. 108468, 2022

    2022

  19. [27]

    FANet: Feature aggregation network for RGBD saliency detection,

    X. Zhou et al., “FANet: Feature aggregation network for RGBD saliency detection,” Signal Processing: Image Communication , vol. 102, p. 116591, 2022

    2022

  20. [28]

    RoadFormer: Duplex Transformer for RGB-Normal se- mantic road scene parsing,

    J. Li et al. , “RoadFormer: Duplex Transformer for RGB-Normal se- mantic road scene parsing,” IEEE Transactions on Intelligent Vehicles , vol. 9, no. 7, pp. 5163–5172, 2024

    2024

  21. [29]

    Roadformer+: Delivering RGB-X scene parsing through scale-aware information decoupling and advanced heterogeneous fea- ture fusion,

    J. Huang et al., “Roadformer+: Delivering RGB-X scene parsing through scale-aware information decoupling and advanced heterogeneous fea- ture fusion,” IEEE Transactions on Intelligent Vehicles , 2024, DOI: 10.1109/TIV .2024.3448251

    2024

  22. [30]

    ViT-CoMer: Vision Transformer with convolutional multi- scale feature interaction for dense predictions,

    C. Xia et al., “ViT-CoMer: Vision Transformer with convolutional multi- scale feature interaction for dense predictions,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 5493–5502

    2024

  23. [31]

    Robust RGB-D fusion for saliency detection,

    Z. Wu et al. , “Robust RGB-D fusion for saliency detection,” in 2022 International Conference on 3D Vision (3DV) . IEEE, 2022, pp. 403– 413

    2022

  24. [32]

    Deep RGB-D saliency detection with depth-sensitive attention and automatic multi-modal fusion,

    P. Sun et al. , “Deep RGB-D saliency detection with depth-sensitive attention and automatic multi-modal fusion,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021, pp. 1407–1417

    2021

  25. [33]

    Gated fully fusion for semantic segmentation,

    X. Li et al., “Gated fully fusion for semantic segmentation,” in Proceed- ings of the AAAI Conference on Artificial Intelligence (AAAI) , vol. 34, no. 07, 2020, pp. 11 418–11 425

    2020

  26. [34]

    UNet++: Redesigning skip connections to exploit mul- tiscale features in image segmentation,

    Z. Zhou et al., “UNet++: Redesigning skip connections to exploit mul- tiscale features in image segmentation,” IEEE Transactions on Medical Imaging, vol. 39, no. 6, pp. 1856–1867, 2019

    2019

  27. [35]

    The CityScapes dataset for semantic urban scene understanding,

    M. Cordts et al. , “The CityScapes dataset for semantic urban scene understanding,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2016, pp. 3213–3223

    2016

  28. [36]

    Object scene flow for autonomous vehicles,

    M. Menze and A. Geiger, “Object scene flow for autonomous vehicles,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2015, pp. 3061–3070

    2015

  29. [37]

    Deep multimodal fusion for semantic image segmen- tation: A survey,

    Y . Zhang et al., “Deep multimodal fusion for semantic image segmen- tation: A survey,” Image and Vision Computing , vol. 105, p. 104042, 2021

    2021

  30. [38]

    AdapNet: Adaptive semantic segmentation in adverse environmental conditions,

    A. Valada et al., “AdapNet: Adaptive semantic segmentation in adverse environmental conditions,” in 2017 IEEE International Conference on Robotics and Automation (ICRA) . IEEE, 2017, pp. 4644–4651

    2017

  31. [39]

    Locality-sensitive deconvolution networks with gated fusion for RGB-D indoor semantic segmentation,

    Y . Cheng et al., “Locality-sensitive deconvolution networks with gated fusion for RGB-D indoor semantic segmentation,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2017, pp. 3029–3037

    2017

  32. [40]

    Learning feature fusion in deep learning-based object detector,

    E. Hassan et al., “Learning feature fusion in deep learning-based object detector,” Journal of Engineering , vol. 2020, no. 1, p. 7286187, 2020

    2020

  33. [41]

    ECFFNet: Effective and consistent feature fusion network for RGB-T salient object detection,

    W. Zhou et al. , “ECFFNet: Effective and consistent feature fusion network for RGB-T salient object detection,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 32, no. 3, pp. 1224– 1235, 2021

    2021

  34. [42]

    Salient object detection: A discriminative regional fea- ture integration approach,

    H. Jiang et al., “Salient object detection: A discriminative regional fea- ture integration approach,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2013, pp. 2083– 2090

    2013

  35. [43]

    ExFuse: Enhancing feature fusion for semantic seg- mentation,

    Z. Zhang et al. , “ExFuse: Enhancing feature fusion for semantic seg- mentation,” in Proceedings of the European Conference on Computer Vision (ECCV), 2018, pp. 269–284

    2018

  36. [44]

    Bi-directional cross-modality feature propagation with separation-and-aggregation gate for RGB-D semantic segmentation,

    X. Chen et al., “Bi-directional cross-modality feature propagation with separation-and-aggregation gate for RGB-D semantic segmentation,” in Proceedings of the European Conference on Computer Vision (ECCV) . Springer, 2020, pp. 561–577

    2020

  37. [45]

    F 3Net: fusion, feedback and focus for salient object de- tection,

    J. Wei et al., “F 3Net: fusion, feedback and focus for salient object de- tection,” in Proceedings of the AAAI conference on Artificial Intelligence (AAAI), vol. 34, no. 07, 2020, pp. 12 321–12 328

    2020

  38. [46]

    Incorporating convolution designs into visual Trans- formers,

    K. Yuan et al. , “Incorporating convolution designs into visual Trans- formers,” in Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , 2021, pp. 579–588

    2021

  39. [47]

    EfficientNet: Rethinking model scaling for con- volutional neural networks,

    M. Tan and Q. Le, “EfficientNet: Rethinking model scaling for con- volutional neural networks,” in International Conference on Machine Learning (ICML). PMLR, 2019, pp. 6105–6114

    2019

  40. [48]

    Deep residual learning for image recognition,

    K. He et al. , “Deep residual learning for image recognition,” in Pro- ceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 770–778

    2016

  41. [49]

    Real-time fusion network for RGB-D semantic seg- mentation incorporating unexpected obstacle detection for road-driving images,

    L. Sun et al. , “Real-time fusion network for RGB-D semantic seg- mentation incorporating unexpected obstacle detection for road-driving images,” IEEE Robotics and Automation Letters, vol. 5, no. 4, pp. 5558– 5565, 2020

    2020

  42. [50]

    THCANet: Two-layer hop cascaded asymptotic network for robot-driving road-scene semantic segmentation in RGB-D images,

    G. Xu et al., “THCANet: Two-layer hop cascaded asymptotic network for robot-driving road-scene semantic segmentation in RGB-D images,” Digital Signal Processing , vol. 136, p. 104011, 2023

    2023

  43. [51]

    Hierarchical dynamic filtering network for RGB-D salient object detection,

    Y . Pang et al. , “Hierarchical dynamic filtering network for RGB-D salient object detection,” in Proceedings of the European Conference on Computer Vision (ECCV) . Springer, 2020, pp. 235–252

    2020

  44. [52]

    Layer normalization,

    J. L. Ba et al. , “Layer normalization,” Computing Research Repository (CoRR) , vol. abs/1607.06450, 2016. [Online]. Available: https://arxiv.org/abs/1607.06450 10

    2016 arXiv

  45. [53]

    Masked-attention mask Transformer for universal image segmentation,

    B. Cheng et al. , “Masked-attention mask Transformer for universal image segmentation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2022, pp. 1290– 1299

    2022

  46. [54]

    FaPN: Feature-aligned pyramid network for dense image prediction,

    S. Huang et al. , “FaPN: Feature-aligned pyramid network for dense image prediction,” in Proceedings of the IEEE/CVF International Con- ference on Computer Vision (ICCV) , 2021, pp. 864–873

    2021

  47. [55]

    Integrating low-level and semantic features for object consistent segmentation,

    H. Fu and G. Qiu, “Integrating low-level and semantic features for object consistent segmentation,” Neurocomputing, vol. 119, pp. 74–81, 2013

    2013

  48. [56]

    Swin Transformer: hierarchical vision Transformer using shifted windows,

    Z. Liu et al. , “Swin Transformer: hierarchical vision Transformer using shifted windows,” in Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , 2021, pp. 10 012–10 022

    2021

  49. [57]

    SegFormer: Simple and efficient design for semantic segmentation with Transformers,

    E. Xie et al. , “SegFormer: Simple and efficient design for semantic segmentation with Transformers,” Advances in Neural Information Pro- cessing Systems (NeurIPS) , vol. 34, pp. 12 077–12 090, 2021

    2021

  50. [58]

    Object-contextual representations for semantic segmen- tation,

    Y . Yuan et al., “Object-contextual representations for semantic segmen- tation,” in Proceedings of the European Conference on Computer Vision (ECCV). Springer, 2020, pp. 173–190

    2020

  51. [59]

    K-Net: Towards unified image segmentation,

    W. Zhang et al. , “K-Net: Towards unified image segmentation,” Ad- vances in Neural Information Processing Systems (NeurIPS) , vol. 34, pp. 10 326–10 338, 2021

    2021

  52. [60]

    Expectation-maximization attention networks for semantic segmentation,

    X. Li et al., “Expectation-maximization attention networks for semantic segmentation,” in Proceedings of the IEEE/CVF International Confer- ence on Computer Vision (ICCV) , 2019, pp. 9167–9176

    2019

  53. [61]

    RAFT-Stereo: Multilevel recurrent field transforms for stereo matching,

    L. Lipson et al., “RAFT-Stereo: Multilevel recurrent field transforms for stereo matching,” in 2021 International Conference on 3D Vision (3DV), 2021, pp. 218–227

    2021

  54. [62]

    Lidar-LLM: Exploring the potential of large language models for 3D lidar understanding,

    S. Yang et al. , “Lidar-LLM: Exploring the potential of large language models for 3D lidar understanding,” Computing Research Repository (CoRR) , vol. abs/2312.14074, 2023. [Online]. Available: https://arxiv.org/abs/2312.14074

    2023 arXiv

  55. [2023]

    Available: https://arxiv.org/abs/2312.14150

    [Online]. Available: https://arxiv.org/abs/2312.14150

    arXiv

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.