Pith. sign in

REVIEW 3 major objections 5 minor 49 references

Modality-Incremental Learning with Disjoint Relevance Mapping Networks for Image-based Semantic Segmentation

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A fixed-size network can add new sensor modalities one at a time without forgetting old ones, by giving each modality its own disjoint set of neural connections.

desk verdict MIL is a useful new framing, but the central DRMN-vs-RMN gain is within seed noise and the 'complete forgetting' claim is overstated. read the letter →

arxiv 2411.17610 v1 pith:ULX2GJJM submitted 2024-11-26 cs.CV

Niharika Hegde , Shishir Muralidhara , René Schuster , Didier Stricker This is my paper · ORCID
classification cs.CV
keywords continuallearningmodality-incrementalsemanticsegmentationrelevancemappingnetworkcatastrophicforgettingmulti-modalperceptionautonomousdrivingparameterisolation
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

Modality-incremental learning (MIL) is a new continual learning setting defined here: a model trained on one sensor modality (say RGB) must later take on new modalities (infrared, depth, grayscale) with the same label set, without forgetting older ones. The paper argues that existing continual learning methods fail here because very different modalities fight for the same weights. It proposes Disjoint Relevance Mapping Networks (DRMN), which extend Relevance Mapping Networks by forcing each task's relevance map to exclude all weights already used by earlier tasks, so every modality learns on its own private set of connections. The central claim is that this hard separation eliminates forgetting and improves new-modality learning compared to shared-weight baselines, while keeping overall network capacity usage nearly unchanged. If true, a single fixed-size network could accommodate an expanding sensor stack in autonomous driving without replaying old data or growing the architecture.

What carries the argument

The central object is the relevance map $M$, a per-task set of weights multiplying the network weights $W$ at each layer, $f_{\text{out}} = W \cdot M_t \cdot f_{\text{in}}$. The original Relevance Mapping Network learns these maps alongside $W$, prunes values below a threshold $\mu$, and freezes the selected important weights for each task. DRMN adds one step: before learning task $t$, every weight that any previous task marked relevant is set to zero in $M_t$, so the new task is restricted to connections no earlier modality used. This is parameter isolation by additive masking: network nodes can still be reused across modalities, but the individual connections cannot. The mechanism is what converts a shared-weight continual learner into a per-modality path allocator, and it also allows all tasks' relevance maps to be stored in one compact structure.

What would settle it

Run each Freiburg Thermal sequence (e.g., RGB→IR→Gray) at least five times with different random seeds for RMN and DRMN and compare the mIoU distributions on the final task; if the DRMN advantage (on average 0.37–1.01 points) is smaller than the typical between-seed spread, the central claim is unsupported. A complementary test: train DRMN on two intentionally similar modalities (RGB then grayscale) and check whether performance drops below RMN, which would show that zero sharing is not universally beneficial.

Watch

Extended reading notes

Core claim

The paper claims that in modality-incremental learning, where each task introduces a new sensor modality with the same set of semantic classes, enforcing a complete separation of relevant neural connections between modalities—disjoint relevance maps—mitigates catastrophic forgetting almost completely without hurting the network's usable capacity. It further claims that weight sharing, which normally aids transfer, becomes detrimental across strongly dissimilar sensing modalities, and that forcing each modality onto previously unused connections produces better final-task accuracy than letting tasks overlap in weight space. The empirical support is a consistent, though small, mIoU advantage of DRMN over the shared-weight RMN baseline on three task orders from Freiburg Thermal (0.37–1.01 points) and on InfraParis (0.10 points), together with per-task evaluations showing performance on earlier modalities stays effectively constant across the learning sequence.

Load-bearing premise

The load-bearing assumption is that the small mIoU advantage of DRMN over RMN reflects a genuine effect of weight separation rather than run-to-run variation, since the paper reports single runs with no error bars.

Editorial extensions

If this is right

  • A fixed-size network can add new sensor modalities sequentially without storing old sensor data or expanding parameters, because each modality leaves a permanent, frozen footprint.
  • Because the sensor itself supplies the task ID at inference time, the network does not need any task-identification or task-inference module.
  • Regularization and distillation methods that rely on shared weights (EWC, ILT) underperform in this setting, suggesting that modality shift is a distinct and more severe form of domain shift.
  • The disjoint masks keep total network utilization almost unchanged relative to RMNs, so capacity is not exhausted faster despite the hard separation; on InfraParis the final task learns from only about 6% of connections and still matches shared-weight RMN.
  • All relevance maps for a growing number of tasks can be kept in a single data structure, so the storage and loading overhead stays constant.

Reading between the lines

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

  • If disjoint masks really eliminate forgetting, then the stability–plasticity dilemma for sensor expansion is resolved by allocation rather than by balancing losses; a natural next test is a similarity-aware overlap that allows partial sharing between modalities that are known to be close (e.g., RGB and grayscale), which the paper itself flags as future work.
  • The result suggests a general recipe for arbitrary input-domain shift, not just sensors: any setting where task identity is known at test time and domains are strongly distinct could use disjoint masks instead of replay or distillation buffers.
  • One testable extension is to pair DRMN with a small rehearsal buffer: if a few stored frames from old modalities are added to new-task training, this should close most of the remaining gap to single-task models without changing the mask structure.
  • The reported 'no forgetting' claim concerns mIoU stability on the tested datasets; measuring per-class and per-pixel forgetting on rare classes (e.g., small traffic objects) would show whether hard separation also protects categories that contribute little to aggregate mIoU.
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

3 major / 5 minor

Summary. The paper introduces modality-incremental learning (MIL), a continual learning setting in which each new task is a new sensor modality for image-based semantic segmentation while the label space stays fixed. The authors propose Disjoint Relevance Mapping Networks (DRMN), a modification of Relevance Mapping Networks (RMN) that masks out connections used by previous tasks, forcing each modality to use disjoint network connections. Experiments on Freiburg Thermal and InfraParis with DeepLabV3+/ResNet-101 compare DRMN against fine-tuning, joint training, EWC, ILT, RMN, and two intermediate variants (ORMN and PRMN). The paper reports that DRMN slightly outperforms RMN on average mIoU across task sequences, keeps prior-task performance stable, and achieves comparable overall network utilization while reducing cross-task weight overlap to zero.

Significance. If the central claim holds, the paper makes a useful contribution: it formalizes a practical continual learning setting (MIL), shows that a fixed-capacity network can accommodate new sensor modalities through parameter isolation, and provides a careful ablation from RMN to ORMN to PRMN to DRMN. The supplementary material is a strength: it reports per-task evolution of mIoU, robustness to the pruning threshold, and detailed network-utilization statistics, which makes the paper's mechanism transparent and its claims checkable. The main weakness is that the decisive DRMN-versus-RMN comparison rests on single-run mIoU differences of 0.10 to 1.08 points, which are within typical seed-to-seed variation for this type of segmentation model. The paper's broad claims about completely mitigating forgetting and about the superiority of disjointness are therefore not yet statistically supported.

major comments (3)
  1. [Sec. 5.2, Tables 1 and 2] The central empirical claim that DRMN improves over RMN in MIL rests on single-run mIoU comparisons. In Table 1, the average DRMN advantage over RMN is 0.37, 0.12, and 1.08 mIoU across the three Freiburg sequences; in Table 2 on InfraParis it is 0.10 mIoU. No error bars, number of seeds, or significance tests are reported anywhere in the paper or supplementary material. For DeepLabV3+ with ResNet-101 trained on such datasets, these gaps are comparable to or smaller than typical run-to-run variation. To support the conclusion that disjointness, rather than optimization noise, drives the improvement, the authors should provide repeated runs (at least three to five seeds) with mean and standard deviation, and ideally a significance test or effect-size estimate, for at least the RMN versus DRMN comparison on all task sequences.
  2. [Supplementary Sec. B, Table 5; Sec. 4.2] The abstract and Sec. 4.2 claim that preventing overlap in relevance maps mitigates forgetting, but the reported data do not show a forgetting advantage for DRMN over RMN. In Table 5, on the sequence IR to Gray to RGB, prior-task mIoU for DRMN is 55.30, 55.16, 54.97 (IR) and for RMN is 55.30, 55.18, 55.10 (IR); both methods are equally stable and have near-zero forgetting. The actual observed advantage of DRMN is on the newly learned modality (e.g., Gray at 70.61 vs. 68.85 and RGB at 71.19 vs. 69.46). Thus the paper should either present explicit forgetting metrics showing that DRMN improves stability relative to RMN, or reframe the contribution as improved forward transfer with equal stability, rather than reduced forgetting.
  3. [Sec. 4.2 and Sec. 6] The method enforces complete disjointness for every modality pair, and the paper claims this is beneficial without hurting capacity. However, all tested modality pairs (RGB, IR, depth, grayscale) are visually dissimilar, and the paper itself concedes in Sec. 4.2 and the conclusion that similar modalities might benefit from shared weights. The claimed advantage of disjointness is therefore only demonstrated for one end of the modality-similarity spectrum. Either experiments with more similar modalities (or with a controlled similarity axis) are needed to support the general claim, or the claim should be explicitly restricted to dissimilar modalities, with the adaptive overlap mechanism left as future work rather than as a caveat.
minor comments (5)
  1. [Sec. 5.2] The GPU name is written as 'RTXA6000'; it should be 'RTX A6000'.
  2. [Eq. (1)] The notation fout = W · Mt · fin would benefit from a brief explanation of how the relevance map multiplies the convolution weights, especially for convolutional layers, since the dimensions of W and Mt are not stated.
  3. [Supplementary Sec. D] In the first sentence, 'network utlization' is a typo for 'network utilization'.
  4. [Reference [4]] The venue is written as 'NeurIOS'; this appears to be a typo for 'NeurIPS'.
  5. [Sec. 5.1] The paper says the Freiburg dataset offers '13 object categories' and InfraParis '20 classes'; it would help to state whether these are the same label sets used for both datasets or whether each dataset has its own evaluation protocol.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: DRMN's claimed benefit is an empirical comparison, not a derivation from its own definitions.

full rationale

The paper's central claim is that enforcing disjoint relevance maps (DRMN) improves modality-incremental learning relative to RMN and other baselines. This is an empirical claim evaluated on held-out datasets (Freiburg Thermal, InfraParis) against external and adapted baselines, not a quantity derived from the method's own definitions. The disjointness itself is enforced by construction in Algorithm 1 (masking previously used connections), but the claimed benefit—higher mIoU on new modalities and stable performance on old ones—is measured, not implied by the definition. The pruning threshold mu = 0.6 is fixed for all methods and its variation is reported as a robustness check in the supplementary material, so no parameter is fitted to the reported mIoU values. The RMN baseline is prior work by Kaushik et al. [25] with no author overlap with the present paper, so no self-citation chain is load-bearing. The only notable weakness is statistical: comparisons rely on single runs without error bars or significance tests, and the DRMN-over-RMN gaps are small (0.10 to 1.01 mIoU). That is a correctness/robustness concern, not circularity. No equation in the paper reduces a prediction to an input by construction, and no load-bearing premise is justified solely by a self-citation.

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

The central claim relies on standard deep learning assumptions (fixed capacity, ImageNet initialization, task ID availability) and on training hyperparameters chosen by hand. No new physical or mathematical entities are postulated, and no constants are fitted to the experimental results.

free parameters (3)
  • Prune threshold mu = 0.6
    Hand-selected threshold for relevance map pruning; robustness across 0.5 and 0.7 is shown in supplementary Table 6, so it is not tightly coupled to results.
  • Prune start epoch = 50
    Chosen by hand; pruning begins from epoch 50 of 75; no sensitivity analysis is reported.
  • Learning rate = 1e-5
    SGD learning rate used for all methods; standard choice, no tuning reported.
assumptions (5)
  • domain assumption Task ID is available at inference time because the sensor that produced the image is known.
    Sec. 3: 'a notable advantage of MIL... task ID... sensor... known to the system.' This enables per-task relevance map selection.
  • domain assumption A single fixed-capacity ResNet-101/DeepLabV3+ network has sufficient disjoint capacity for up to four modalities.
    Sec. 4.2 and Table 7: DRMN utilizes 6.37% of connections for the final InfraParis task, so capacity is not exhausted in the tested settings; this may not hold for larger modality sets.
  • standard math The RMN formulation f_out = W * M_t * f_in and the pruning/freezing procedure from Kaushik et al. are taken as given.
    Sec. 4.1, Eq. 1; the paper extends RMN without re-deriving its properties.
  • domain assumption All modalities share the same label space Y.
    Sec. 3: MIL requires 'the set of classes Y remains consistent across all tasks, similar to DIL'.
  • domain assumption Pretrained ImageNet weights provide a suitable initialization for all modalities.
    Sec. 5.2: DeepLabV3+ with ResNet-101 pretrained on ImageNet is used for all models; no modality-specific pretraining.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Modality-Incremental Learning with Disjoint Relevance Mapping Networks for Image-based Semantic Segmentation." pith.science (2026). https://pith.science/paper/ULX2GJJM

@misc{pith2026241117610,
  author       = {Pith},
  title        = {Pith review of: Modality-Incremental Learning with Disjoint Relevance Mapping Networks for Image-based Semantic Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ULX2GJJM}},
  note         = {Machine review of arXiv:2411.17610}
}
read the original abstract

In autonomous driving, environment perception has significantly advanced with the utilization of deep learning techniques for diverse sensors such as cameras, depth sensors, or infrared sensors. The diversity in the sensor stack increases the safety and contributes to robustness against adverse weather and lighting conditions. However, the variance in data acquired from different sensors poses challenges. In the context of continual learning (CL), incremental learning is especially challenging for considerably large domain shifts, e.g. different sensor modalities. This amplifies the problem of catastrophic forgetting. To address this issue, we formulate the concept of modality-incremental learning and examine its necessity, by contrasting it with existing incremental learning paradigms. We propose the use of a modified Relevance Mapping Network (RMN) to incrementally learn new modalities while preserving performance on previously learned modalities, in which relevance maps are disjoint. Experimental results demonstrate that the prevention of shared connections in this approach helps alleviate the problem of forgetting within the constraints of a strict continual learning framework.

Figures

Figures reproduced from arXiv: 2411.17610 by the authors.

Figure 1
Figure 1. Three different modalities to perceive traffic scenarios in [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Relevance Mapping Network (RMN) (left) shares connections across tasks, with new tasks utilizing their respective relevance [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. A Relevance Mapping Network augments the network [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 38 canonical work pages

  1. [1]

    Memory aware synapses: Learning what (not) to forget

    Rahaf Aljundi, Francesca Babiloni, Mohamed Elhoseiny, Marcus Rohrbach, and Tinne Tuytelaars. Memory aware synapses: Learning what (not) to forget. In ECCV, 2018. 2

    2018

  2. [2]

    Continual Road-Scene Semantic Segmentation via Feature-Aligned Symmetric Multi-Modal Network

    Francesco Barbato, Elena Camuffo, Simone Milani, and Pietro Zanuttigh. Continual road-scene semantic segmen- tation via feature-aligned symmetric multi-modal network. arXiv preprint arXiv:2308.04702, 2023. 3

    work page Pith review arXiv 2023

  3. [3]

    Modeling the background for incremental learning in semantic segmentation

    Fabio Cermelli, Massimiliano Mancini, Samuel Rota Bul `o, Elisa Ricci, and Barbara Caputo. Modeling the background for incremental learning in semantic segmentation. InCVPR,

    work page

  4. [4]

    Ssul: Semantic segmentation with unknown label for exemplar- based class-incremental learning

    Sungmin Cha, YoungJoon Yoo, Taesup Moon, et al. Ssul: Semantic segmentation with unknown label for exemplar- based class-incremental learning. Advances in Neural In- formation Processing Systems (NeurIOS), 2021. 2

    work page 2021

  5. [5]

    Encoder-decoder with atrous separable convolution for semantic image segmenta- tion

    Liang-Chieh Chen, Yukun Zhu, George Papandreou, Flo- rian Schroff, and Hartwig Adam. Encoder-decoder with atrous separable convolution for semantic image segmenta- tion. arXiv:1802.02611, 2018. 6

    arXiv 2018

  6. [6]

    Bi-directional cross-modality feature propagation with separation-and- aggregation gate for rgb-d semantic segmentation

    Xiaokang Chen, Kwan-Yee Lin, Jingbo Wang, Wayne Wu, Chen Qian, Hongsheng Li, and Gang Zeng. Bi-directional cross-modality feature propagation with separation-and- aggregation gate for rgb-d semantic segmentation. In ECCV,

    work page

  7. [7]

    Indoor semantic segmentation using depth in- formation: 1st international conference on learning represen- tations, iclr 2013

    Camille Couprie, Cl ´ement Farabet, Laurent Najman, and Yann LeCun. Indoor semantic segmentation using depth in- formation: 1st international conference on learning represen- tations, iclr 2013. In 1st International Conference on Learn- ing Representations, ICLR 2013, 2013. 3

    work page 2013

  8. [8]

    RFBNet: Deep Multimodal Networks with Residual Fusion Blocks for RGB-D Semantic Segmentation

    Liuyuan Deng, Ming Yang, Tianyi Li, Yuesheng He, and Chunxiang Wang. Rfbnet: deep multimodal networks with residual fusion blocks for rgb-d semantic segmentation. arXiv preprint arXiv:1907.00135, 2019. 3

    work page Pith review arXiv 1907

Show all 49 references
  1. [9]

    Replaying styles for continual semantic segmentation across domains

    Yao Deng and Xiang Xiang. Replaying styles for continual semantic segmentation across domains. In Pattern Recogni- tion: Asian Conference, 2023. 2

    2023

  2. [10]

    Plop: Learning without forgetting for con- tinual semantic segmentation

    Arthur Douillard, Yifu Chen, Arnaud Dapogny, and Matthieu Cord. Plop: Learning without forgetting for con- tinual semantic segmentation. In CVPR, 2021. 2

    2021

  3. [11]

    Pathnet: Evolution channels gradient descent in super neural networks

    Chrisantha Fernando, Dylan Banarse, Charles Blundell, Yori Zwols, David Ha, Andrei A Rusu, Alexander Pritzel, and Daan Wierstra. Pathnet: Evolution channels gradient descent in super neural networks. arXiv preprint arXiv:1701.08734,

    arXiv

  4. [12]

    Infraparis: A multi-modal and multi-task autonomous driving dataset

    Gianni Franchi, Marwane Hariat, Xuanlong Yu, Nacim Belkhir, Antoine Manzanera, and David Filliat. Infraparis: A multi-modal and multi-task autonomous driving dataset. In WACV, 2024. 3, 5, 7, 8, 11, 12

    2024

  5. [13]

    Born again neural net- works

    Tommaso Furlanello, Zachary Lipton, Michael Tschannen, Laurent Itti, and Anima Anandkumar. Born again neural net- works. In ICML, 2018. 2

    2018

  6. [14]

    Multi-domain incremental learning for semantic segmenta- tion

    Prachi Garg, Rohit Saluja, Vineeth N Balasubramanian, Chetan Arora, Anbumani Subramanian, and CV Jawahar. Multi-domain incremental learning for semantic segmenta- tion. In WACV, 2022. 2

    2022

  7. [15]

    A bio-inspired in- cremental learning architecture for applied perceptual prob- lems

    Alexander Gepperth and Cem Karaoguz. A bio-inspired in- cremental learning architecture for applied perceptual prob- lems. Cognitive Computation, 2016. 2

    2016

  8. [16]

    Attribution-aware weight transfer: A warm- start initialization for class-incremental semantic segmenta- tion

    Dipam Goswami, Ren ´e Schuster, Joost van de Weijer, and Didier Stricker. Attribution-aware weight transfer: A warm- start initialization for class-incremental semantic segmenta- tion. In WACV, 2023. 2

    2023

  9. [17]

    Memory efficient experience replay for streaming learning

    Tyler L Hayes, Nathan D Cahill, and Christopher Kanan. Memory efficient experience replay for streaming learning. In International Conference on Robotics and Automation (ICRA), 2019. 2

    2019

  10. [18]

    Fusenet: Incorporating depth into semantic seg- mentation via fusion-based cnn architecture

    Caner Hazirbas, Lingni Ma, Csaba Domokos, and Daniel Cremers. Fusenet: Incorporating depth into semantic seg- mentation via fusion-based cnn architecture. InACCV, 2017. 3

    2017

  11. [19]

    Deep residual learning for image recognition

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

  12. [20]

    Overcoming catastrophic interfer- ence by conceptors

    Xu He and Herbert Jaeger. Overcoming catastrophic interfer- ence by conceptors. arXiv preprint arXiv:1707.04853, 2017. 2

    2017 arXiv

  13. [21]

    Learning a unified classifier incrementally via rebalancing

    Saihui Hou, Xinyu Pan, Chen Change Loy, Zilei Wang, and Dahua Lin. Learning a unified classifier incrementally via rebalancing. In CVPR, 2019. 2

    2019

  14. [22]

    Acnet: Attention based network to exploit complementary features for rgbd semantic segmentation

    Xinxin Hu, Kailun Yang, Lei Fei, and Kaiwei Wang. Acnet: Attention based network to exploit complementary features for rgbd semantic segmentation. In ICIP, 2019. 3

    2019

  15. [23]

    Using conceptors to manage neural long- term memories for temporal patterns

    Herbert Jaeger. Using conceptors to manage neural long- term memories for temporal patterns. Journal of Machine Learning Research, 2017. 2

    2017

  16. [24]

    Less-forgetting learning in deep neural networks

    Heechul Jung, Jeongwoo Ju, Minju Jung, and Junmo Kim. Less-forgetting learning in deep neural networks. arXiv preprint arXiv:1607.00122, 2016. 2

    2016 arXiv

  17. [25]

    Understanding catastrophic forgetting and remem- bering in continual learning with optimal relevance mapping

    Prakhar Kaushik, Alex Gain, Adam Kortylewski, and Alan Yuille. Understanding catastrophic forgetting and remem- bering in continual learning with optimal relevance mapping. arXiv preprint arXiv:2102.11343, 2021. 2, 4, 6, 7, 8, 11, 12

    2021 arXiv

  18. [26]

    Fearnet: Brain- inspired model for incremental learning

    Ronald Kemker and Christopher Kanan. Fearnet: Brain- inspired model for incremental learning. arXiv preprint arXiv:1711.10563, 2017. 2

    2017 arXiv

  19. [27]

    Overcoming catastrophic forgetting in neu- ral networks

    James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska- Barwinska, et al. Overcoming catastrophic forgetting in neu- ral networks. Proceedings of the national academy of sc...

    2017

  20. [28]

    Packnet: Adding mul- tiple tasks to a single network by iterative pruning

    Arun Mallya and Svetlana Lazebnik. Packnet: Adding mul- tiple tasks to a single network by iterative pruning. InCVPR,

  21. [29]

    Catastrophic inter- ference in connectionist networks: The sequential learning problem

    Michael McCloskey and Neal J Cohen. Catastrophic inter- ference in connectionist networks: The sequential learning problem. In Psychology of Learning and Motivation . Else- vier, 1989. 1

    1989

  22. [30]

    The stability-plasticity dilemma: Investigating the contin- uum from catastrophic forgetting to age-limited learning ef- fects

    Martial Mermillod, Aur ´elia Bugaiska, and Patrick Bonin. The stability-plasticity dilemma: Investigating the contin- uum from catastrophic forgetting to age-limited learning ef- fects. Frontiers in Psychology, 2013. 1

    2013

  23. [31]

    Knowledge dis- tillation for incremental learning in semantic segmentation

    Umberto Michieli and Pietro Zanuttigh. Knowledge dis- tillation for incremental learning in semantic segmentation. Computer Vision and Image Understanding , 2021. 2, 6, 7, 11, 12

    2021

  24. [32]

    Motion and depth augmented semantic segmentation for autonomous navigation

    Hazem Rashed, Ahmad El Sallab, Senthil Yogamani, and Mohamed ElHelw. Motion and depth augmented semantic segmentation for autonomous navigation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2019. 3

    2019

  25. [33]

    Towards domain-aware knowledge distillation for continual model generalization

    Nikhil Reddy, Mahsa Baktashmotlagh, and Chetan Arora. Towards domain-aware knowledge distillation for continual model generalization. In WACV, 2024. 2

    2024

  26. [34]

    Berg, and Li Fei-Fei

    Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, San- jeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, Alexander C. Berg, and Li Fei-Fei. ImageNet Large Scale Visual Recognition Chal- lenge. International Journal of Computer Vision (IJCV) ,

  27. [35]

    Progressive neural networks

    Andrei A Rusu, Neil C Rabinowitz, Guillaume Desjardins, Hubert Soyer, James Kirkpatrick, Koray Kavukcuoglu, Raz- van Pascanu, and Raia Hadsell. Progressive neural networks. arXiv preprint arXiv:1606.04671, 2016. 2

    2016 arXiv

  28. [36]

    Continual learning with deep generative replay

    Hanul Shin, Jung Kwon Lee, Jaehong Kim, and Jiwon Kim. Continual learning with deep generative replay. NeurIPS,

  29. [37]

    Rtfnet: Rgb- thermal fusion network for semantic segmentation of urban scenes

    Yuxiang Sun, Weixun Zuo, and Ming Liu. Rtfnet: Rgb- thermal fusion network for semantic segmentation of urban scenes. IEEE Robotics and Automation Letters, 2019. 3

    2019

  30. [38]

    Fuseseg: Semantic segmentation of urban scenes based on rgb and thermal data fusion

    Yuxiang Sun, Weixun Zuo, Peng Yun, Hengli Wang, and Ming Liu. Fuseseg: Semantic segmentation of urban scenes based on rgb and thermal data fusion. IEEE Transactions on Automation Science and Engineering, 2021. 3

    2021

  31. [39]

    Heatnet: Bridging the day-night domain gap in semantic segmenta- tion with thermal images

    Johan Vertens, Jannik Z ¨urn, and Wolfram Burgard. Heatnet: Bridging the day-night domain gap in semantic segmenta- tion with thermal images. arXiv preprint arXiv:2003.04645,

    2003 arXiv

  32. [40]

    Learn- ing deep multimodal feature representation with asymmetric multi-layer fusion

    Yikai Wang, Fuchun Sun, Ming Lu, and Anbang Yao. Learn- ing deep multimodal feature representation with asymmetric multi-layer fusion. In Proceedings of the 28th ACM Interna- tional Conference on Multimedia, 2020. 3

    2020

  33. [41]

    Grow- ing a brain: Fine-tuning by increasing model capacity

    Yu-Xiong Wang, Deva Ramanan, and Martial Hebert. Grow- ing a brain: Fine-tuning by increasing model capacity. In CVPR, 2017. 2

    2017

  34. [42]

    Memory replay gans: Learn- ing to generate new categories without forgetting

    Chenshen Wu, Luis Herranz, Xialei Liu, Joost Van De Wei- jer, Bogdan Raducanu, et al. Memory replay gans: Learn- ing to generate new categories without forgetting. NeurIPS,

  35. [43]

    Ccaffmnet: Dual-spectral semantic segmentation network with channel- coordinate attention feature fusion module

    Shi Yi, Junjie Li, Xi Liu, and Xuesong Yuan. Ccaffmnet: Dual-spectral semantic segmentation network with channel- coordinate attention feature fusion module. Neurocomput- ing, 2022. 3

    2022

  36. [44]

    Lifelong learning with dynamically expandable net- works

    Jaehong Yoon, Eunho Yang, Jeongtae Lee, and Sung Ju Hwang. Lifelong learning with dynamically expandable net- works. In ICLR, 2018. 2

    2018

  37. [45]

    Contin- ual learning through synaptic intelligence

    Friedemann Zenke, Ben Poole, and Surya Ganguli. Contin- ual learning through synaptic intelligence. In ICML, 2017. 2

    2017

  38. [46]

    Representation compensation networks for continual semantic segmentation

    Chang-Bin Zhang, Jia-Wen Xiao, Xialei Liu, Ying-Cong Chen, and Ming-Ming Cheng. Representation compensation networks for continual semantic segmentation. In CVPR,

  39. [47]

    Cmx: Cross-modal fusion for rgb-x semantic segmentation with transformers

    Jiaming Zhang, Huayao Liu, Kailun Yang, Xinxin Hu, Ruip- ing Liu, and Rainer Stiefelhagen. Cmx: Cross-modal fusion for rgb-x semantic segmentation with transformers. IEEE Transactions on Intelligent Transportation Systems, 2023. 3

    2023

  40. [48]

    Abmdrnet: Adaptive-weighted bi-directional modality difference reduc- tion network for rgb-t semantic segmentation

    Qiang Zhang, Shenlu Zhao, Yongjiang Luo, Dingwen Zhang, Nianchang Huang, and Jungong Han. Abmdrnet: Adaptive-weighted bi-directional modality difference reduc- tion network for rgb-t semantic segmentation. In CVPR,

  41. [49]

    Gmnet: Graded-feature multilabel-learning network for rgb-thermal urban scene semantic segmentation

    Wujie Zhou, Jinfu Liu, Jingsheng Lei, Lu Yu, and Jenq-Neng Hwang. Gmnet: Graded-feature multilabel-learning network for rgb-thermal urban scene semantic segmentation. IEEE Transactions on Image Processing, 2021. 3 Supplementary Material A. Overview In this supplementary materi...

    2021

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.