Pith. sign in

REVIEW 5 major objections 6 minor 57 references

Optimizing Dense Visual Predictions Through Multi-Task Coherence and Prioritization

T0 review · 5 major / 6 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 MT-CP reaches new bests on NYUD-v2 and PASCAL-Context with trace-back and loss prioritization.

desk verdict Big reported gains on MTL benchmarks, but the loss-prioritization mechanism is over-sold; still deserves a serious referee. read the letter →

arxiv 2412.03179 v1 pith:7EQYT6KS submitted 2024-12-04 cs.CV cs.AI

Maxime Fontana , Michael Spratling , Miaojing Shi This is my paper · ORCID
classification cs.CVcs.AI
keywords multi-tasklearningdensepredictioncross-taskcoherencetrace-backlossprioritizationsemanticsegmentationdepthestimationsurfacenormal
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

MT-CP is a multi-task learning model for dense visual prediction that claims new best results on the NYUD-v2 and PASCAL-Context benchmarks. The paper's central idea is that multi-task performance improves when tasks stay coherent in both geometry and prediction, and when their losses are balanced dynamically without learned parameters. To that end it introduces a trace-back procedure that fuses cross-task representations and refines initial predictions through task-specific decoders, plus a Loss Prioritization Scheme that projects losses to a common scale and up-weights tasks that are improving slowly. On NYUD-v2 the full model reports semantic segmentation mIoU of 56.25, depth RMSE of 0.4316, and normal error of 18.60; on PASCAL-Context it reports semseg mIoU 79.96, parsing mIoU 69.13, and saliency maxF 84.20. Ablations show that the coherence fusion module is the largest single contributor, with the loss scheme adding a further gain over equal weighting and plain log smoothing.

What carries the argument

The load-bearing mechanism is the pairing of a Coherence Fusion Module (CFM) and a Spatial Refinement Module (SRM), together called Trace-Back, with a Loss Prioritization Scheme (LPS). CFM enforces geometric coherence by aligning task representations with cosine distance and merging them by matrix multiplication; SRM then propagates the fused representation back through the task decoder's stages and keeps intermediate predictions in the loss. LPS computes task weights from historical loss ratios, $\tilde{w}_i^n = \frac{\prod_{k=1}^H L_i^{n-k+1}/L_i^{n-k}}{\prod_{k=1}^H L^{n-k+1}/L^{n-k}}$, applies them as $\log(1+w_i)L_i$ to project all losses onto a common scale, and uses a spread parameter $\kappa$ to widen or narrow the weight differences. These three pieces are what the ablations isolate.

What would settle it

Record per-task gradient norms on the shared backbone during NYUD-v2 training under LPS. The scale-projection premise predicts that reweighted task gradients have comparable magnitudes; if one task's gradient norm still dominates by an order of magnitude while the reported metric gains persist, the proposed mechanism is not what is driving the improvement.

Watch

Extended reading notes

Core claim

The paper claims that the combination of cross-task coherence and parameter-free loss prioritization is what drives dense multi-task performance. Coherence is enforced twice: a Coherence Fusion Module aligns each task's representation with the fused auxiliary representations by minimizing cosine distance and merging them through matrix multiplication, and a Spatial Refinement Module traces the fused representation back through the task decoder's stages, adding intermediate predictions to the loss. Balancing is handled by the Loss Prioritization Scheme, which defines the total loss as $\sum_i \log(1+w_i)L_i$ and updates $w_i$ each epoch from the ratio of task-loss decreases over a history window, with a spread parameter $\kappa$ controlling how far weights deviate from their mean. The reported numbers on NYUD-v2 (56.25 mIoU, 0.4316 RMSE, 18.60 mErr) and PASCAL-Context (79.96 mIoU, 69.13 mIoU, 84.20 maxF) are the evidence offered for the claim.

Load-bearing premise

The load-bearing premise is that multiplying each task loss by $\log(1+w_i)$, with $w_i$ computed from historical loss ratios, really does project all task losses onto a common scale; if that premise fails, the reported gains over plain log smoothing and equal weighting may reflect other training details rather than the prioritization scheme.

Editorial extensions

If this is right

  • If MT-CP is correct, the same CFM/SRM architecture should transfer to other dense task sets, such as adding edge or optical-flow heads, without changing the coherence or prioritization machinery.
  • The Loss Prioritization Scheme is parameter-free, so it can be dropped into other multi-task models; the paper's comparisons suggest it would beat equal weighting and plain log smoothing there as well.
  • The ablation results imply that removing CFM costs more than removing SRM, and removing both costs most; a practitioner building a cheaper model can drop SRM first with less damage.
  • The method's reported gains over MLoRE and Bi-MTDP position it as the new comparison point for dense multi-task prediction on NYUD-v2 and PASCAL-Context.

Reading between the lines

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

  • A reader could test the loss-prioritization claim more sharply by updating weights at a finer granularity than once per epoch, since the paper's Eq. (4) only sees epoch-level loss ratios.
  • Because the backbone is pretrained on ADE20K, the absolute scores partly reflect that pretraining; swapping in a different backbone would isolate how much of the gain comes from CFM, SRM, and LPS rather than from the shared encoder.
  • The cosine-distance alignment in CFM is a form of feature-space regularization, which suggests the module might also help single-task models that want geometric consistency across scales, though the paper does not test that.
  • The spread parameter $\kappa$ is global; an adaptive per-task version could be explored, since the paper's own Figure 6 shows weight trajectories differing across tasks.
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

5 major / 6 minor

Summary. The paper proposes MT-CP, a multi-task learning model for dense visual prediction tasks. The architecture combines a Mask2Former backbone with task-specific transformer decoders, a Coherence Fusion Module (CFM) that aligns task representations through gating, cosine distance, and matrix multiplication, and Spatial Refinement Modules (SRMs) that trace cross-task representations back through the decoder stages to refine predictions. The third contribution is a dynamic Loss Prioritization Scheme (LPS) that reweights task losses using historical loss-decrease ratios, a log transform, and a spread parameter. Experiments on NYUD-v2 and PASCAL-Context report state-of-the-art or competitive results on semantic segmentation, depth estimation, surface normal estimation, part parsing, and saliency detection. Ablations on NYUD-v2 show that the architectural components and the LPS each contribute to the reported performance.

Significance. If the reported results are reproducible, MT-CP is a useful empirical contribution: it improves depth RMSE on NYUD-v2 from 0.5076 to 0.4316 over MLoRE, and it is competitive on PASCAL-Context, with code publicly released. The architecture is reasonably simple and the component-wise ablations (Table 3) give some evidence that CFM and SRM help. The main weakness is that the paper's central methodological claim about LPS—that multiplying task losses by log(1+w_i) projects losses onto a common scale—is not supported by the equations, and the evaluation has confounds (backbone differences, single runs, ablations only on one dataset). These issues need to be resolved before the state-of-the-art claims can be accepted.

major comments (5)
  1. [Sec. 3.5, Eq. (3)-(4)] The claim that Eq. (3) 'has the effect to remove the scale imbalance problem' is not supported. The weights w_i defined in Eq. (4) are products of ratios of relative loss decreases, L_i^{n-k+1}/L_i^{n-k}, divided by the analogous global-loss ratios. These quantities depend only on the relative speed of loss reduction, not on the absolute magnitude of each task loss. Therefore, if task A has a raw loss ten times larger than task B but both follow the same relative decrease trajectory, Eq. (4) assigns them equal weights and Eq. (3) preserves the ten-fold scale gap. The ablation in Table 4 therefore cannot isolate the 'common scale' effect claimed. Please provide a derivation of the scale-projection property, or better, a controlled experiment in which the weighting rule is applied to explicitly normalized losses so that the scale effect is separated from the prioritization effect.
  2. [Sec. 3.5, Eq. (6)] Equation (6) can produce negative weights under the reported setting kappa=2.5. Since w_i' = mu_i + kappa(wi - mu_i), a task with wi < mu_i - mu_i/kappa (for kappa=2.5, wi < 0.6 mu_i) receives a negative w_i'. The weights w_i are ratios of loss-decrease products, so such values are plausible. Then log(1+w_i') in Eq. (3) is either negative or undefined, directly contradicting the statement that the log transform 'is necessary to avoid values for w_i in [0,1] leading negative weights.' The update rule needs a non-negativity constraint or a proof that w_i' stays positive, and the statement in Sec. 3.5 must be revised accordingly.
  3. [Sec. 4.2, 4.3, Tables 1-2] The comparison with prior state-of-the-art methods is potentially confounded by the backbone and pretraining. MT-CP uses a Mask2Former backbone pre-trained on ADE20K for semantic segmentation, while typical baselines such as MLoRE, Bi-MTDP, and InvPT may use different backbones or pretraining procedures. The paper does not report the backbone and pretraining for each baseline, nor does it provide a same-backbone comparison. Given that some gains are small (e.g., 55.96 to 56.25 mIoU on NYUD-v2 semantic segmentation), the reader cannot distinguish the contribution of the proposed MTL method from the advantage of the stronger backbone. Please specify the backbones of all compared methods or add a same-backbone baseline.
  4. [Table 4] The loss-scheme ablation that supports the LPS contribution is reported only on NYUD-v2. The paper claims that LPS is central to the state-of-the-art results on both NYUD-v2 and PASCAL-Context, but no LPS ablation is shown on PASCAL-Context. Moreover, the reported numbers in Tables 1-4 appear to come from single training runs; for gains of 0.1-0.3 mIoU, this is not sufficient to establish that the differences are not due to training noise. Please add an LPS ablation on PASCAL-Context and report means and standard deviations over at least three seeds for the key comparisons.
  5. [Sec. 1 and Sec. 3.5] The text in the introduction and abstract describes LPS as 'parameter-free,' but the method has two hyperparameters, kappa and history length H, whose values are selected empirically (kappa=2.5, H=3 in Sec. 4.2). The only sensitivity analysis is Fig. 6, which qualitatively shows the effect of kappa but does not report the corresponding task metrics or any sensitivity to H. Please either remove the 'parameter-free' claim or clarify that it means 'no learned parameters,' and report a sensitivity study for both kappa and H.
minor comments (6)
  1. [Sec. 3.5, Eq. (7)] The notation 's.t. w* = LPS(w, kappa)' in Eq. (7) is not defined; the optimization problem is not formalized. Please state explicitly how the weights are computed and how they enter the loss, or rewrite Eq. (7) as a plain sum with a defined update rule.
  2. [Sec. 3.3] The text says 'we minimize the cosine distance between XT1' and XTaux,' but no cosine-distance loss term appears in the overall MTL loss in Eq. (7). Please clarify whether the cosine distance is a training loss, a regularizer, or an architectural property, and if it is a loss, show it explicitly in the total objective.
  3. [Sec. 3.5, first paragraph] The sentence 'we denote Li to be the loss for a task i in a set T according to Eq. (3)' is confusing because Eq. (3) defines a weighted sum over all tasks, not a per-task loss. Please use a distinct symbol for the raw per-task loss and indicate which loss enters Eq. (4) and Eq. (7).
  4. [Sec. 4.4, Table 4] The row 'MT-CP (w/ Log Smoothing)' should define exactly what 'log smoothing' means: is it Eq. (3) with equal weights w_i=1/T, or with the LPS weights but without the spread update? Without this definition, the comparison between 'Log Smoothing' and 'Loss Prioritization' is not interpretable.
  5. [Sec. 4.4, paragraph after Table 3] There is a typo: 'though less severely.suggesting' should read 'though less severely, suggesting'.
  6. [References] References [28] and [36] appear to be the same paper (Silberman et al., Indoor Segmentation and Support Inference from RGBD Images) with different author order and formatting; please consolidate. Also, several references lack venue or arXiv IDs (e.g., [10], [18], [19]) and the GitHub URL in the abstract has a missing space before the colon.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: benchmark results are measured externally, LPS weights come from training dynamics rather than test labels, and the self-citations are not load-bearing. The Eq. (3) scale-projection gap is a correctness concern, not circularity.

full rationale

The paper's central claims are evaluated on external benchmarks (NYUD-v2, PASCAL-Context), and the proposed LPS weights in Eq. (4) are computed from ratios of historical training losses, not from test labels or from the final reported metrics. The loss objective Eq. (3) is therefore a training heuristic whose performance is independently measured, so the SOTA numbers are not forced by construction. Architectural components (CFM, SRM/Trace-Back) are ablated directly on the benchmark metrics, and the ablations do not define the reported numbers. Self-citations [7,57] appear only in background discussion and do not carry the load-bearing argument; no uniqueness theorem or prior-work-only justification is invoked. The most plausible circularity-adjacent concern is substantive but not circular: the text says Eq. (3) 'remove[s] the scale imbalance problem,' yet the weights from Eq. (4) encode relative loss-decrease ratios, not absolute loss magnitudes, so the claimed scale projection does not obviously follow from the equations. This is a mathematical-mechanism gap that could be tested by additional ablations or gradient-scale analysis, but it does not make the benchmark predictions equivalent to the method's inputs. Hyperparameters kappa and H are selected on validation data, which is standard practice rather than fitting-to-prediction. No step in the derivation chain reduces to an input by definition, so the circularity score is low.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim rests on several design assumptions validated only by the paper's own experiments: geometric coherence via cosine alignment, matrix-multiplication projection, task difficulty via loss ratios, and a log-weight scaling asserted to equalize loss scales. The external anchors are the two public benchmarks and prior SOTA tables; no code commit or checkpoints are provided to independently verify the reported scores.

free parameters (3)
  • Kappa (spread parameter) = 2.5
    Controls the spread of task weights around the mean in Eq. (6). Chosen empirically in Sec. 4.2; Fig. 6 shows Kappa affects training dynamics, and no automatic selection or sensitivity range is provided.
  • H (history length) = 3
    Window length for the loss-ratio products in Eq. (4). Chosen empirically in Sec. 4.2 with no sensitivity analysis.
  • Task decoder block configuration = (2,2,2) blocks, depth (1,2,1)
    Architectural hyperparameters of the task-specific decoders (Sec. 4.2); not ablated and chosen by hand.
assumptions (5)
  • domain assumption Cosine distance between task feature vectors enforces geometric coherence.
    Invoked in Sec. 3.3 (CFM) to justify aligning feature directions. No theoretical or empirical demonstration that this preserves scene geometry across tasks.
  • domain assumption Matrix multiplication of task features projects them into a common space and maintains consistency.
    Also in Sec. 3.3; the projection and consistency claims are asserted without derivation.
  • ad hoc to paper Task difficulty can be measured by the product of historical loss-decrease ratios (Eq. 4), and prioritizing slower tasks improves MTL.
    This is the paper's heuristic for LPS. It is not derived from an optimization principle and depends on the empirically chosen H and Kappa.
  • ad hoc to paper Multiplying each task loss by log(1+w_i) projects losses onto a common scale and removes scale imbalance.
    Stated with Eq. (3) in Sec. 3.5. This is mathematically unsupported: a per-task constant factor does not equalize losses with different intrinsic units and magnitudes.
  • domain assumption A Mask2Former backbone pre-trained on ADE20K transfers effectively to NYUD-v2 and PASCAL-Context.
    Sec. 4.2: all reported results inherit the quality and domain of this pretrained backbone; its contribution is not isolated by an ablation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Optimizing Dense Visual Predictions Through Multi-Task Coherence and Prioritization." pith.science (2026). https://pith.science/paper/7EQYT6KS

@misc{pith2026241203179,
  author       = {Pith},
  title        = {Pith review of: Optimizing Dense Visual Predictions Through Multi-Task Coherence and Prioritization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7EQYT6KS}},
  note         = {Machine review of arXiv:2412.03179}
}
read the original abstract

Multi-Task Learning (MTL) involves the concurrent training of multiple tasks, offering notable advantages for dense prediction tasks in computer vision. MTL not only reduces training and inference time as opposed to having multiple single-task models, but also enhances task accuracy through the interaction of multiple tasks. However, existing methods face limitations. They often rely on suboptimal cross-task interactions, resulting in task-specific predictions with poor geometric and predictive coherence. In addition, many approaches use inadequate loss weighting strategies, which do not address the inherent variability in task evolution during training. To overcome these challenges, we propose an advanced MTL model specifically designed for dense vision tasks. Our model leverages state-of-the-art vision transformers with task-specific decoders. To enhance cross-task coherence, we introduce a trace-back method that improves both cross-task geometric and predictive features. Furthermore, we present a novel dynamic task balancing approach that projects task losses onto a common scale and prioritizes more challenging tasks during training. Extensive experiments demonstrate the superiority of our method, establishing new state-of-the-art performance across two benchmark datasets. The code is available at:https://github.com/Klodivio355/MT-CP

Figures

Figures reproduced from arXiv: 2412.03179 by the authors.

Figure 1
Figure 1. Our MTL framework implements cross-task coherence [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The proposed MT-CP model. Only two tasks are shown for clarity. The model consists of a shared set of features extracted by [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The coherence fusion module. use the notation XT ′ 1 and XT′ aux to describe the resulting representations. Then, we design two strategies to enforce geometric coherence to help enhance the main task. Firstly, we minimize the cosine distance between XT ′ 1 and XT′ aux , the cosine distance ensures that the vectors in each repre￾sentation are attracted together towards the same direction. This conceptually helps ensu… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: The spatial refinement module used to trace back cross [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Visualisations of predictions on NYUD-v2 [ [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Variation of the spread value κ on our Loss Prioritization Scheme (LPS). 5. Conclusion This paper introduces MT-CP, a multi-task learning model designed for dense prediction tasks. MT-CP effec￾tively leverages pixel-wise cross-task information through each task-specifi…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 51 canonical work pages

  1. [1]

    Multimae: Multi-modal multi-task masked autoen- coders, 2022

    Roman Bachmann, David Mizrahi, Andrei Atanov, and Amir Zamir. Multimae: Multi-modal multi-task masked autoen- coders, 2022. 1

    work page 2022

  2. [2]

    Exploring rela- tional context for multi-task dense prediction, 2021

    David Bruggemann, Menelaos Kanakis, Anton Obukhov, Stamatios Georgoulis, and Luc Van Gool. Exploring rela- tional context for multi-task dense prediction, 2021. 2

    work page 2021

  3. [3]

    Gradnorm: Gradient normalization for adaptive loss balancing in deep multitask networks

    Zhao Chen, Vijay Badrinarayanan, Chen-Yu Lee, and An- drew Rabinovich. Gradnorm: Gradient normalization for adaptive loss balancing in deep multitask networks. 2017. 3

    work page 2017

  4. [4]

    Schwing, Alexan- der Kirillov, and Rohit Girdhar

    Bowen Cheng, Ishan Misra, Alexander G. Schwing, Alexan- der Kirillov, and Rohit Girdhar. Masked-attention mask transformer for universal image segmentation. In 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 1280–1289, 2022. 3, 4, 6

    work page 2022

  5. [5]

    Se- mantic image segmentation: Two decades of research, 2023

    Gabriela Csurka, Riccardo V olpi, and Boris Chidlovskii. Se- mantic image segmentation: Two decades of research, 2023. 1

    work page 2023

  6. [6]

    Everingham, L

    M. Everingham, L. Van Gool, C. K. I. Williams, J. Winn, and A. Zisserman. The pascal visual object classes (voc) chal- lenge. International Journal of Computer Vision, 88(2):303– 338, June 2010. 2, 6, 7

    work page 2010

  7. [7]

    When multi-task learning meets partial supervision: A com- puter vision review, 2024

    Maxime Fontana, Michael Spratling, and Miaojing Shi. When multi-task learning meets partial supervision: A com- puter vision review, 2024. 1

    work page 2024

  8. [8]

    Yuan Gao, Qi She, Jiayi Ma, Mingbo Zhao, Wei Liu, and Alan L. Yuille. NDDR-CNN: layer-wise feature fusing in multi-task CNN by neural discriminative dimensionality re- duction. CoRR, abs/1801.08297, 2018. 1, 2

    work page Pith review arXiv 2018

Show all 57 references
  1. [9]

    R-cnns for pose estimation and action detec- tion, 2014

    Georgia Gkioxari, Bharath Hariharan, Ross Girshick, and Ji- tendra Malik. R-cnns for pose estimation and action detec- tion, 2014. 3

    2014

  2. [10]

    Dynamic task prioritization for multitask learning

    Michelle Guo, Albert Haque, De-An Huang, Serena Yeung, and Li Fei-Fei. Dynamic task prioritization for multitask learning. In Proceedings of the European Conference on Computer Vision (ECCV), September 2018. 3

    2018

  3. [11]

    Unit: Multimodal mul- titask learning with a unified transformer, 2021

    Ronghang Hu and Amanpreet Singh. Unit: Multimodal mul- titask learning with a unified transformer, 2021. 2

    2021

  4. [12]

    Lau, and Thomas S

    Jianbo Jiao, Yunchao Wei, Zequn Jie, Honghui Shi, Ryn- son W.H. Lau, and Thomas S. Huang. Geometry-aware dis- tillation for indoor semantic segmentation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2019. 1, 2, 4

    2019

  5. [13]

    Multi-task learning using uncertainty to weigh losses for scene geome- try and semantics

    Alex Kendall, Yarin Gal, and Roberto Cipolla. Multi-task learning using uncertainty to weigh losses for scene geome- try and semantics. CoRR, abs/1705.07115, 2017. 2, 3, 5

    2017 arXiv

  6. [14]

    Pushing the boundaries of boundary de- tection using deep learning, 2016

    Iasonas Kokkinos. Pushing the boundaries of boundary de- tection using deep learning, 2016. 1

    2016

  7. [15]

    Ubernet: Training a ’universal’ convo- lutional neural network for low-, mid-, and high-level vi- sion using diverse datasets and limited memory

    Iasonas Kokkinos. Ubernet: Training a ’universal’ convo- lutional neural network for low-, mid-, and high-level vi- sion using diverse datasets and limited memory. CoRR, abs/1609.02132, 2016. 1, 2, 3, 5

    2016 arXiv

  8. [16]

    Learning multi- ple pixelwise tasks based on loss scale balancing

    Jae-Han Lee, Chul Lee, and Chang-Su Kim. Learning multi- ple pixelwise tasks based on loss scale balancing. In 2021 IEEE/CVF International Conference on Computer Vision (ICCV), pages 5087–5096, 2021. 3

    2021

  9. [17]

    Transformed dynamic feature pyramid for small object detection

    Hong Liang, Ying Yang, Qian Zhang, Linxia Feng, Jie Ren, and Qiyao Liang. Transformed dynamic feature pyramid for small object detection. IEEE Access, PP:1–1, 09 2021. 4

    2021

  10. [18]

    Auxiliary tasks in multi- task learning, 2018

    Lukas Liebel and Marco K ¨orner. Auxiliary tasks in multi- task learning, 2018. 2, 5

    2018

  11. [19]

    Baijiong Lin, Weisen Jiang, Feiyang Ye, Yu Zhang, Peng- guang Chen, Ying-Cong Chen, Shu Liu, and James T. Kwok. Dual-balancing for multi-task learning, 2023. 3, 5

    2023

  12. [20]

    Rethinking boundary detection in deep learning models for medical image segmentation, 2023

    Yi Lin, Dong Zhang, Xiao Fang, Yufan Chen, Kwang-Ting Cheng, and Hao Chen. Rethinking boundary detection in deep learning models for medical image segmentation, 2023. 1

    2023

  13. [21]

    Shikun Liu, Edward Johns, and Andrew J. Davison. End-to-end multi-task learning with attention. CoRR, abs/1803.10704, 2018. 2

    2018 arXiv

  14. [22]

    Swin trans- former: Hierarchical vision transformer using shifted win- dows, 2021

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin trans- former: Hierarchical vision transformer using shifted win- dows, 2021. 4, 6

    2021

  15. [23]

    Cross- task attention mechanism for dense multi-task learning,

    Ivan Lopes, Tuan-Hung Vu, and Raoul de Charette. Cross- task attention mechanism for dense multi-task learning,

  16. [24]

    Decoupled weight decay regularization, 2019

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization, 2019. 6

    2019

  17. [25]

    Jiaqi Ma, Zhe Zhao, Xinyang Yi, Jilin Chen, Lichan Hong, and Ed H. Chi. Modeling task relationships in multi-task learning with multi-gate mixture-of-experts. Proceedings of the 24th ACM SIGKDD International Conference on Knowl- edge Discovery & Data Mining, 2018. 2

    2018

  18. [26]

    Image seg- mentation using deep learning: A survey, 2020

    Shervin Minaee, Yuri Boykov, Fatih Porikli, Antonio Plaza, Nasser Kehtarnavaz, and Demetri Terzopoulos. Image seg- mentation using deep learning: A survey, 2020. 1

    2020

  19. [27]

    Cross-stitch networks for multi-task learning

    Ishan Misra, Abhinav Shrivastava, Abhinav Gupta, and Mar- tial Hebert. Cross-stitch networks for multi-task learning. CoRR, abs/1604.03539, 2016. 1, 2, 6

    2016 arXiv

  20. [28]

    Indoor segmentation and support inference from rgbd images

    Pushmeet Kohli Nathan Silberman, Derek Hoiem and Rob Fergus. Indoor segmentation and support inference from rgbd images. In ECCV, 2012. 2, 5, 6, 7, 8

    2012

  21. [29]

    An overview of multi-task learning in deep neural networks, 2017

    Sebastian Ruder. An overview of multi-task learning in deep neural networks, 2017. 1

    2017

  22. [30]

    Latent multi-task architecture learning,

    Sebastian Ruder, Joachim Bingel, Isabelle Augenstein, and Anders Søgaard. Latent multi-task architecture learning,

  23. [31]

    Rusu, Neil C

    Andrei A. Rusu, Neil C. Rabinowitz, Guillaume Desjardins, Hubert Soyer, James Kirkpatrick, Koray Kavukcuoglu, Raz- van Pascanu, and Raia Hadsell. Progressive neural networks,

  24. [32]

    Saurabh Saxena, Abhishek Kar, Mohammad Norouzi, and David J. Fleet. Monocular depth estimation using diffusion models, 2023. 1

    2023

  25. [33]

    Efficient multitask dense predictor via binarization, 2024

    Yuzhang Shang, Dan Xu, Gaowen Liu, Ramana Rao Kom- pella, and Yan Yan. Efficient multitask dense predictor via binarization, 2024. 6, 7

    2024

  26. [34]

    Learning to multi-task by active sampling,

    Sahil Sharma, Ashutosh Jha, Parikshit Hegde, and Balara- man Ravindran. Learning to multi-task by active sampling,

  27. [35]

    Usb: Universal-scale object detection bench- mark, 2021

    Yosuke Shinya. Usb: Universal-scale object detection bench- mark, 2021. 2

    2021

  28. [36]

    Indoor segmentation and support inference from rgbd images

    Nathan Silberman, Derek Hoiem, Pushmeet Kohli, and Rob Fergus. Indoor segmentation and support inference from rgbd images. volume 7576, pages 746–760, 10 2012. 2

    2012

  29. [37]

    Training data-efficient image transformers and distillation through at- tention, 2021

    Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, and Herv ´e J´egou. Training data-efficient image transformers and distillation through at- tention, 2021. 4

    2021

  30. [38]

    Multi-task learning for dense prediction tasks: A sur- vey

    Simon Vandenhende, Stamatios Georgoulis, Wouter Van Gansbeke, Marc Proesmans, Dengxin Dai, and Luc Van Gool. Multi-task learning for dense prediction tasks: A sur- vey. IEEE Transactions on Pattern Analysis and Machine Intelligence, pages 1–1, 2021. 1

    2021

  31. [39]

    Mti-net: Multi-scale task interaction networks for multi-task learning, 2020

    Simon Vandenhende, Stamatios Georgoulis, and Luc Van Gool. Mti-net: Multi-scale task interaction networks for multi-task learning, 2020. 2, 6

    2020

  32. [40]

    Internim- age: Exploring large-scale vision foundation models with deformable convolutions, 2022

    Wenhai Wang, Jifeng Dai, Zhe Chen, Zhenhang Huang, Zhiqi Li, Xizhou Zhu, Xiaowei Hu, Tong Lu, Lewei Lu, Hongsheng Li, Xiaogang Wang, and Yu Qiao. Internim- age: Exploring large-scale vision foundation models with deformable convolutions, 2022. 1, 2

    2022

  33. [41]

    Pyra- mid vision transformer: A versatile backbone for dense pre- diction without convolutions, 2021

    Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo, and Ling Shao. Pyra- mid vision transformer: A versatile backbone for dense pre- diction without convolutions, 2021. 4

    2021

  34. [42]

    PVT v2: Improved baselines with pyramid vision transformer

    Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo, and Ling Shao. PVT v2: Improved baselines with pyramid vision transformer. Computational Visual Media, 8(3):415–424, mar 2022. 4

    2022

  35. [43]

    Gradient vaccine: Investigating and improving multi-task optimization in massively multilingual models

    Zirui Wang, Yulia Tsvetkov, Orhan Firat, and Yuan Cao. Gradient vaccine: Investigating and improving multi-task optimization in massively multilingual models. In Interna- tional Conference on Learning Representations, 2021. 3

    2021

  36. [44]

    Cbam: Convolutional block attention module, 2018

    Sanghyun Woo, Jongchan Park, Joon-Young Lee, and In So Kweon. Cbam: Convolutional block attention module, 2018. 4

    2018

  37. [45]

    Pad-net: Multi-tasks guided prediction-and-distillation net- work for simultaneous depth estimation and scene parsing,

    Dan Xu, Wanli Ouyang, Xiaogang Wang, and Nicu Sebe. Pad-net: Multi-tasks guided prediction-and-distillation net- work for simultaneous depth estimation and scene parsing,

  38. [46]

    Mtformer: Multi-task learning via transformer and cross-task reasoning

    Xiaogang Xu, Hengshuang Zhao, Vibhav Vineet, Ser-Nam Lim, and Antonio Torralba. Mtformer: Multi-task learning via transformer and cross-task reasoning. In Computer Vi- sion – ECCV 2022: 17th European Conference, Tel Aviv, Is- rael, October 23–27, 2022, Proceedings, Part XXVII ...

    2022

  39. [47]

    Demt: De- formable mixer transformer for multi-task learning of dense prediction, 2023

    Yangyang Xu, Yibo Yang, and Lefei Zhang. Demt: De- formable mixer transformer for multi-task learning of dense prediction, 2023. 6

    2023

  40. [48]

    Depth anything: Unleashing the power of large-scale unlabeled data, 2024

    Lihe Yang, Bingyi Kang, Zilong Huang, Xiaogang Xu, Jiashi Feng, and Hengshuang Zhao. Depth anything: Unleashing the power of large-scale unlabeled data, 2024. 1, 2

    2024

  41. [49]

    Multi-task dense prediction via mixture of low-rank experts, 2024

    Yuqi Yang, Peng-Tao Jiang, Qibin Hou, Hao Zhang, Jinwei Chen, and Bo Li. Multi-task dense prediction via mixture of low-rank experts, 2024. 6, 7

    2024

  42. [50]

    Invpt: Inverted pyramid multi-task transformer for dense scene understanding

    Hanrong Ye and Dan Xu. Invpt: Inverted pyramid multi-task transformer for dense scene understanding. 2022. 6

    2022

  43. [51]

    Taskprompter: Spatial-channel multi-task prompting for dense scene understanding

    Hanrong Ye and Dan Xu. Taskprompter: Spatial-channel multi-task prompting for dense scene understanding. In The Eleventh International Conference on Learning Representa- tions, 2023. 1, 2, 3, 5

    2023

  44. [52]

    Gradient surgery for multi-task learning, 2020

    Tianhe Yu, Saurabh Kumar, Abhishek Gupta, Sergey Levine, Karol Hausman, and Chelsea Finn. Gradient surgery for multi-task learning, 2020. 3

    2020

  45. [53]

    A survey on multi-task learn- ing

    Yu Zhang and Qiang Yang. A survey on multi-task learn- ing. IEEE Transactions on Knowledge and Data Engineer- ing, 34(12):5586–5609, 2022. 1

    2022

  46. [54]

    Pattern-affinitive propagation across depth, surface normal and semantic segmentation, 2019

    Zhenyu Zhang, Zhen Cui, Chunyan Xu, Yan Yan, Nicu Sebe, and Jian Yang. Pattern-affinitive propagation across depth, surface normal and semantic segmentation, 2019. 2, 6

    2019

  47. [55]

    Semantic under- standing of scenes through the ade20k dataset, 2018

    Bolei Zhou, Hang Zhao, Xavier Puig, Tete Xiao, Sanja Fi- dler, Adela Barriuso, and Antonio Torralba. Semantic under- standing of scenes through the ade20k dataset, 2018. 6

    2018

  48. [56]

    L. Zhou, Z. Cui, C. Xu, Z. Zhang, C. Wang, T. Zhang, and J. Yang. Pattern-structure diffusion for multi-task learning. In 2020 IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR), pages 4513–4522, Los Alamitos, CA, USA, jun 2020. IEEE Computer Society. 6

    2020

  49. [57]

    Vlprompt: Vision-language prompting for panoptic scene graph gener- ation, 2024

    Zijian Zhou, Miaojing Shi, and Holger Caesar. Vlprompt: Vision-language prompting for panoptic scene graph gener- ation, 2024. 2

    2024

Pith tools

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