Pith. sign in

REVIEW 1 major objections 1 minor 53 references

A hierarchical transformer fuses RGB and event data to reduce fusion noise in object detection.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.3

2026-06-30 08:05 UTC pith:C4TB66M4

load-bearing objection CMTFormer gives a staged RGB-event fusion design with SAM/CEM/LDFM but the mAP gains do not isolate whether those modules actually cut noise or redundancy. the 1 major comments →

arxiv 2606.29136 v1 pith:C4TB66M4 submitted 2026-06-28 cs.CV cs.AI

CMTFormer: Marrying Transformer with Hierarchical Information Interaction for RGB-Event Object Detection

Yu Li , Yuenan Hou , Yingmei Wei , Jiangming Chen , Yanming Guo This is my paper
classification cs.CV cs.AI
keywords RGB-event fusionobject detectiontransformermultimodal collaborationevent camerahierarchical interactioncross-modal enhancement
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proposes CMTFormer to address the challenge of fusing RGB frames with event camera streams, which have different attributes that can cause noise or redundancy in previous methods. It introduces a shallow-to-deep interaction scheme with three modules: SAM for low-level alignment, CEM for middle-level mutual enhancement using texture and edges, and LDFM for high-level learnable fusion. A Spatial Prior Module aids localization. Experiments on two benchmarks show it outperforms other detectors in both single and combined modality settings. This matters because effective multimodal fusion could leverage event cameras' high temporal resolution and dynamic range for more robust detection.

Core claim

CMTFormer marries a transformer architecture with hierarchical cross-modal information interaction, using SAM to fuse low-level features while mitigating disparities, CEM to reinforce middle-level features with texture and edge info, LDFM to aggregate high-level clues via learnable weights, and a Spatial Prior Module for better localization, resulting in superior performance on RGB-event object detection tasks.

What carries the argument

The shallow-to-deep information interaction scheme with Shallow Alignment Module (SAM), Cross-modal Enhancement Module (CEM), and Learnable Deep Fusion Module (LDFM) that enables efficient and stable multimodal collaboration.

Load-bearing premise

The shallow-to-deep interaction scheme mitigates attribute disparities and prevents noisy or redundant integration between RGB and event data without introducing new fusion artifacts.

What would settle it

An ablation study showing that removing the hierarchical modules (SAM, CEM, LDFM) does not degrade performance on the DSEC-Detection benchmark would falsify the central claim.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Surpasses detection counterparts in both uni-modal and multi-modal settings on DSEC-Detection and PKU-DAVIS-SOD benchmarks.
  • Prevents noisy or redundant feature integration during cross-modal fusion.
  • The learnable weights in LDFM allow adaptive fusion of RGB and event clues.
  • Global spatial information from the Spatial Prior Module enhances localization accuracy.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar hierarchical fusion could be applied to other sensor combinations like RGB and LiDAR.
  • The approach may generalize to video object detection where temporal information is key.
  • By avoiding new fusion artifacts, it could enable more reliable deployment in challenging environments like low light or fast motion.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

1 major / 1 minor

Summary. The paper proposes CMTFormer, a transformer architecture for RGB-event object detection that employs a shallow-to-deep hierarchical interaction scheme. This includes the Shallow Alignment Module (SAM) for low-level fusion to reduce attribute disparities, the Cross-modal Enhancement Module (CEM) for mutual reinforcement of middle-level features using texture and edge cues, the Learnable Deep Fusion Module (LDFM) for adaptive high-level aggregation via learnable weights, and an additional Spatial Prior Module for improved localization. The central claim is that this design enables efficient multimodal collaboration without noise amplification or redundancy, leading to consistent outperformance over uni-modal and multi-modal baselines on the DSEC-Detection and PKU-DAVIS-SOD benchmarks.

Significance. If the empirical results hold under rigorous validation, the hierarchical fusion paradigm could advance RGB-event detection by addressing modality heterogeneity more systematically than prior fusion techniques. The learnable adaptive fusion in LDFM and the staged interaction represent a structured approach that may generalize to other event-based vision tasks, with the promise of reproducibility noted via planned code release.

major comments (1)
  1. [Experiments section] Experiments section (and associated tables/figures): The claim that SAM, CEM, and LDFM specifically mitigate attribute disparities and prevent noisy/redundant integration (as stated in the abstract and §1) rests solely on overall mAP improvements versus baselines. No ablation studies isolating each module's contribution, nor direct quantitative analyses (e.g., feature correlation, noise statistics, or redundancy metrics before/after fusion), are described to confirm the modules achieve the stated mitigation rather than other factors such as transformer capacity or training details.
minor comments (1)
  1. [Abstract] Abstract: No quantitative metrics, baseline names, or specific mAP values are reported, which weakens the standalone readability of the claims.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive comment on the experimental validation of our proposed modules. We address the point below and will revise the manuscript accordingly to strengthen the evidence.

read point-by-point responses

  1. Referee: [Experiments section] Experiments section (and associated tables/figures): The claim that SAM, CEM, and LDFM specifically mitigate attribute disparities and prevent noisy/redundant integration (as stated in the abstract and §1) rests solely on overall mAP improvements versus baselines. No ablation studies isolating each module's contribution, nor direct quantitative analyses (e.g., feature correlation, noise statistics, or redundancy metrics before/after fusion), are described to confirm the modules achieve the stated mitigation rather than other factors such as transformer capacity or training details.

    Authors: We agree that the current experiments demonstrate overall mAP gains but do not isolate the specific contributions of SAM, CEM, and LDFM through dedicated ablations or direct metrics on disparity mitigation, noise, or redundancy. While the staged design rationale (detailed in §3) and consistent outperformance over comparable baselines provide supporting context, additional targeted experiments are needed to rule out confounding factors such as model capacity. In the revised manuscript, we will add: (i) module-wise ablation tables reporting performance when each component is removed individually, (ii) capacity-controlled variants that retain transformer blocks but omit the hierarchical cross-modal interaction, and (iii) supplementary quantitative analyses (e.g., feature correlation or similarity metrics across modalities before/after fusion) together with qualitative visualizations. These additions will directly address the concern. revision: yes

Circularity Check

0 steps flagged

No circularity; empirical architecture validated on external benchmarks

full rationale

The paper proposes CMTFormer with descriptive modules (SAM, CEM, LDFM, Spatial Prior) for RGB-event fusion and supports effectiveness solely via end-to-end mAP gains on DSEC-Detection and PKU-DAVIS-SOD against uni- and multi-modal baselines. No equations, derivations, fitted parameters renamed as predictions, or load-bearing self-citations appear. The central claims reduce to experimental comparison rather than any self-referential construction, making the result self-contained against external data.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no mathematical derivations, parameters, or assumptions are stated in the provided text.

pith-pipeline@v0.9.1-grok · 5811 in / 975 out tokens · 33039 ms · 2026-06-30T08:05:32.803489+00:00 · methodology

0 comments
read the original abstract

Event cameras capture sparse brightness changes with high temporal resolution and high dynamic range, compensating for the deficiencies of the conventional RGB frames. However, previous multi-modal fusion techniques typically fail to handle the inherent heterogeneity between RGB frames and event streams, thus easily leading to noise amplification or redundant feature integration during cross-modal fusion. In this paper, we propose a Cross-Modal information inTeraction transFormer, coined as CMTFormer, which hierarchically integrates RGB and event information to achieve efficient and stable multimodal collaboration. Specifically, we design a shallow-to-deep information interaction scheme. In the shallow stage, we present the Shallow Alignment Module (SAM) to achieve an efficient fusion of RGB and event low-level features, which mitigates attribute disparities and prevents noisy information. In the middle stage, we devise the Cross-modal Enhancement Module (CEM) that utilizes texture and edge information to produce mutually reinforced middle-level features. In the deep stage, we present the Learnable Deep Fusion Module (LDFM) which performs high-level information aggregation through learnable weights, thus enabling the network to adaptively fuse RGB and event clues. A Spatial Prior Module is further designed to utilize global spatial information to enhance localization accuracy. Extensive experiments are conducted on two prevalent event-based object detection benchmarks, i.e., DSEC-Detection and PKU-DAVIS-SOD. Our CMTFormer consistently surpasses the detection counterparts in both uni-modal and multi-modal settings, strongly demonstrating the effectiveness of our paradigm. Codes will be available upon publication.

Figures

Figures reproduced from arXiv: 2606.29136 by Jiangming Chen, Yanming Guo, Yingmei Wei, Yuenan Hou, Yu Li.

Figure 1
Figure 1. Figure 1: Schematic overview of our CMTFormer. (a) The architec [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Framework overview of our CMTFormer. Our model takes stacked event frames and RGB frames as inputs and extracts multi-scale [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Illustration of the Shallow Alignment Module (SAM), a [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: The architecture of the Learnable Deep Fusion Module [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Overview of the Spatial-Prior Module, which uses the global [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Visual predictions of CMTFormer under di [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Visual comparison of CMTFormer with state-of-the-art methods on the DSEC-Detection dataset [ [PITH_FULL_IMAGE:figures/full_fig_p012_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Visualization of asynchronous detections produced by com [PITH_FULL_IMAGE:figures/full_fig_p013_9.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

53 extracted references · 53 canonical work pages · 3 internal anchors

  1. [1]

    Z. Zou, K. Chen, Z. Shi, Y . Guo, J. Ye, Object detection in 20 years: A survey, Proceedings of the IEEE 111 (3) (2023) 257–276

    work page 2023

  2. [2]

    Zhou, A yolo-nl object detector for real-time detection, Expert Systems with Applications 238 (2024) 122256

    Y . Zhou, A yolo-nl object detector for real-time detection, Expert Systems with Applications 238 (2024) 122256

    work page 2024

  3. [3]

    Y . Hou, Z. Ma, C. Liu, C. C. Loy, Learning lightweight lane detection CNNs by self attention distillation, in: IEEE International Conference on Computer Vision, 2019, pp. 1013–1021

    work page 2019

  4. [4]

    Y . Ma, T. Wang, X. Bai, H. Yang, Y . Hou, Y . Wang, Y . Qiao, R. Yang, X. Zhu, Vision-centric bev perception: A survey, IEEE Transactions on Pattern Analysis and Machine Intelligence 46 (12) (2024) 10978–10997

    work page 2024

  5. [5]

    W. Li, W. Yang, Y . Hou, L. Liu, Y . Liu, X. Li, Saratr-x: Toward building a foundation model for sar target recognition, IEEE Transactions on Im- age Processing 34 (2025) 869–884

    work page 2025

  6. [6]

    Y . Li, Y . Hou, Y . Wei, X. Zhu, Y . Ma, W. Shao, Y . Guo, Moe3d: Mixture of experts meets multi-modal 3d understanding, arXiv preprint arXiv:2511.22103 (2025)

    work page arXiv 2025

  7. [7]

    Y . Hu, T. Delbruck, S.-C. Liu, Learning to ex- ploit multiple vision modalities by using grafted networks, in: European Conference on Computer Vision, Springer, 2020, pp. 85–101

    work page 2020

  8. [8]

    Chakravarthi, A

    B. Chakravarthi, A. A. Verma, K. Daniilidis, C. Fermuller, Y . Yang, Recent event camera inno- vations: A survey, in: European Conference on Computer Vision, Springer, 2024, pp. 342–376

    work page 2024

  9. [9]

    D. Liu, Y . Fan, W. Lu, C. Liu, W. Zhang, Spiking depth: Depth estimation from sparse events with spiking neural networks, Expert Systems with Ap- plications (2025) 129977

    work page 2025

  10. [10]

    D. Kang, D. Kang, An adaptive learning frame- work for event-based remote eye tracking, Expert Systems with Applications (2025) 128038

    work page 2025

  11. [11]

    H. Cao, G. Chen, J. Xia, G. Zhuang, A. Knoll, Fusion-based feature attention gate component for vehicle detection based on event camera, IEEE Sensors Journal 21 (21) (2021) 24540–24548

    work page 2021

  12. [12]

    A. Tomy, A. Paigwar, K. S. Mann, A. Renzaglia, C. Laugier, Fusing event-based and rgb camera for robust object detection in adverse conditions, in: 2022 International Conference on Robotics and Automation (ICRA), IEEE, 2022, pp. 933–939

    work page 2022

  13. [13]

    D. Li, Y . Tian, J. Li, Sodformer: Streaming ob- ject detection with transformer using events and frames, IEEE Transactions on Pattern Analysis and Machine Intelligence 45 (11) (2023) 14020– 14037

    work page 2023

  14. [14]

    Jiang, Y

    Y . Jiang, Y . Wang, M. Zhao, Y . Zhang, H. Qi, Nighttime traffic object detection via adaptively integrating event and frame domains, Fundamen- tal Research (2023)

    work page 2023

  15. [15]

    J. Li, S. Dong, Z. Yu, Y . Tian, T. Huang, Event- based vision enhanced: A joint detection frame- work in autonomous driving, in: 2019 ieee interna- tional conference on multimedia and expo (icme), IEEE, 2019, pp. 1396–1401

    work page 2019

  16. [16]

    Jiang, P

    Z. Jiang, P. Xia, K. Huang, W. Stechele, G. Chen, Z. Bing, A. Knoll, Mixed frame-/event-driven fast pedestrian detection, in: 2019 International Conference on Robotics and Automation (ICRA), IEEE, 2019, pp. 8332–8338

    work page 2019

  17. [17]

    D. Li, J. Li, X. Liu, Z. Zhou, X. Fan, Y . Tian, Hdi-former: Hybrid dynamic interaction ann-snn 13 transformer for object detection using frames and events, arXiv preprint arXiv:2411.18658 (2024)

    work page arXiv 2024

  18. [19]

    Gehrig, D

    D. Gehrig, D. Scaramuzza, Low-latency automo- tive vision with event cameras, Nature 629 (8014) (2024) 1034–1040

    work page 2024

  19. [20]

    Y . Peng, Y . Zhang, P. Xiao, X. Sun, F. Wu, Better and faster: Adaptive event conversion for event-based object detection, in: Proceedings of the AAAI Conference on Artificial Intelligence, V ol. 37, 2023, pp. 2056–2064

    work page 2023

  20. [21]

    Gehrig, D

    M. Gehrig, D. Scaramuzza, Recurrent vision transformers for object detection with event cam- eras, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 13884–13893

    work page 2023

  21. [22]

    Bodden, D

    L. Bodden, D. B. Ha, F. Schwaiger, L. Kreuzberg, S. Behnke, Spiking centernet: A distillation- boosted spiking neural network for object detec- tion, in: 2024 International Joint Conference on Neural Networks (IJCNN), IEEE, 2024, pp. 1–9

    work page 2024

  22. [23]

    Schaefer, D

    S. Schaefer, D. Gehrig, D. Scaramuzza, Aegnn: Asynchronous event-based graph neural networks, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 12371–12381

    work page 2022

  23. [24]

    Chitta, A

    K. Chitta, A. Prakash, B. Jaeger, Z. Yu, K. Renz, A. Geiger, Transfuser: Imitation with transformer- based sensor fusion for autonomous driving, IEEE transactions on pattern analysis and machine intel- ligence 45 (11) (2022) 12878–12895

    work page 2022

  24. [25]

    Tulyakov, D

    S. Tulyakov, D. Gehrig, S. Georgoulis, J. Erbach, M. Gehrig, Y . Li, D. Scaramuzza, Time lens: Event-based video frame interpolation, in: Pro- ceedings of the IEEE/CVF conference on com- puter vision and pattern recognition, 2021, pp. 16155–16164

    work page 2021

  25. [26]

    P. Duan, Z. W. Wang, B. Shi, O. Cossairt, T. Huang, A. K. Katsaggelos, Guided event filter- ing: Synergy between intensity images and neu- romorphic events for high performance imaging, IEEE Transactions on Pattern Analysis and Ma- chine Intelligence 44 (11) (2021) 8261–8275

    work page 2021

  26. [27]

    Zhang, X

    J. Zhang, X. Yang, Y . Fu, X. Wei, B. Yin, B. Dong, Object tracking by jointly exploiting frame and event domain, in: Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 13043–13052

    work page 2021

  27. [28]

    Gehrig, M

    D. Gehrig, M. Rüegg, M. Gehrig, J. Hidalgo- Carrió, D. Scaramuzza, Combining events and frames using recurrent asynchronous multimodal networks for monocular depth prediction, IEEE Robotics and Automation Letters 6 (2) (2021) 2822–2829

    work page 2021

  28. [29]

    Y .-F. Zuo, J. Yang, J. Chen, X. Wang, Y . Wang, L. Kneip, Devo: Depth-event camera visual odom- etry in challenging conditions, in: 2022 Interna- tional Conference on Robotics and Automation (ICRA), IEEE, 2022, pp. 2179–2185

    work page 2022

  29. [30]

    L. Gao, Y . Liang, J. Yang, S. Wu, C. Wang, J. Chen, L. Kneip, Vector: A versatile event- centric benchmark for multi-sensor slam, IEEE Robotics and Automation Letters 7 (3) (2022) 8217–8224

    work page 2022

  30. [31]

    M. Liu, N. Qi, Y . Shi, B. Yin, An attention fusion network for event-based vehicle object detection, in: 2021 IEEE International Conference on Image Processing (ICIP), IEEE, 2021, pp. 3363–3367

    work page 2021

  31. [32]

    L. Fan, J. Yang, L. Wang, J. Zhang, X. Lian, H. Shen, Efficient spiking neural network for rgb– event fusion-based object detection, Electronics 14 (6) (2025) 1105

    work page 2025

  32. [33]

    R. L. Graham, Concrete mathematics: a founda- tion for computer science, Pearson Education In- dia, 1994

    work page 1994

  33. [34]

    A. Z. Zhu, L. Yuan, K. Chaney, K. Daniilidis, Un- supervised event-based learning of optical flow, depth, and egomotion, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 989–997

    work page 2019

  34. [35]

    N. F. Chen, Pseudo-labels for supervised learning on dynamic vision sensor data, applied to object detection under ego-motion, in: Proceedings of the IEEE conference on computer vision and pattern recognition workshops, 2018, pp. 644–653

    work page 2018

  35. [36]

    A. I. Maqueda, A. Loquercio, G. Gallego, N. Gar- cía, D. Scaramuzza, Event-based vision meets 14 deep learning on steering prediction for self- driving cars, in: Proceedings of the IEEE confer- ence on computer vision and pattern recognition, 2018, pp. 5419–5427

    work page 2018

  36. [37]

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

    work page 2016

  37. [38]

    X. Zhu, W. Su, L. Lu, B. Li, X. Wang, J. Dai, Deformable detr: Deformable transform- ers for end-to-end object detection, arXiv preprint arXiv:2010.04159 (2020)

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  38. [39]

    Kinga, J

    D. Kinga, J. B. Adam, et al., A method for stochas- tic optimization, in: International conference on learning representations (ICLR), V ol. 5, Califor- nia;, 2015

    work page 2015

  39. [40]

    T.-Y . Lin, M. Maire, S. Belongie, J. Hays, P. Per- ona, D. Ramanan, P. Dollár, C. L. Zitnick, Mi- crosoft coco: Common objects in context, in: Eu- ropean conference on computer vision, Springer, 2014, pp. 740–755

    work page 2014

  40. [41]

    S. Ren, K. He, R. Girshick, J. Sun, Faster r-cnn: Towards real-time object detection with region proposal networks, IEEE transactions on pattern analysis and machine intelligence 39 (6) (2016) 1137–1149

    work page 2016

  41. [42]

    T.-Y . Lin, P. Goyal, R. Girshick, K. He, P. Dollár, Focal loss for dense object detection, in: Proceed- ings of the IEEE international conference on com- puter vision, 2017, pp. 2980–2988

    work page 2017

  42. [43]

    X. Zhou, D. Wang, P. Krähenbühl, Objects as points, arXiv preprint arXiv:1904.07850 (2019)

    work page internal anchor Pith review Pith/arXiv arXiv 1904

  43. [44]

    X. Zhu, S. Lyu, X. Wang, Q. Zhao, Tph-yolov5: Improved yolov5 based on transformer prediction head for object detection on drone-captured sce- narios, in: Proceedings of the IEEE/CVF interna- tional conference on computer vision, 2021, pp. 2778–2788

    work page 2021

  44. [45]

    C.-Y . Wang, A. Bochkovskiy, H.-Y . M. Liao, Yolov7: Trainable bag-of-freebies sets new state- of-the-art for real-time object detectors, in: Pro- ceedings of the IEEE/CVF conference on com- puter vision and pattern recognition, 2023, pp. 7464–7475

    work page 2023

  45. [46]

    Z. Liu, H. Mao, C.-Y . Wu, C. Feichtenhofer, T. Darrell, S. Xie, A convnet for the 2020s, in: Proceedings of the IEEE/CVF conference on com- puter vision and pattern recognition, 2022, pp. 11976–11986

    work page 2022

  46. [47]

    Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, B. Guo, Swin transformer: Hierarchical vi- sion transformer using shifted windows, in: Pro- ceedings of the IEEE/CVF international confer- ence on computer vision, 2021, pp. 10012–10022

    work page 2021

  47. [48]

    Q. Su, Y . Chou, Y . Hu, J. Li, S. Mei, Z. Zhang, G. Li, Deep directly-trained spiking neural net- works for object detection, in: Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 6555–6565

    work page 2023

  48. [49]

    Z. Liu, N. Yang, Y . Wang, Y . Li, X. Zhao, F.- Y . Wang, Enhancing traffic object detection in variable illumination with rgb-event fusion, IEEE Transactions on Intelligent Transportation Sys- tems (2024)

    work page 2024

  49. [50]

    H. Cao, Z. Zhang, Y . Xia, X. Li, J. Xia, G. Chen, A. Knoll, Embracing events and frames with hier- archical feature refinement network for object de- tection, in: European Conference on Computer Vi- sion, Springer, 2024, pp. 161–177

    work page 2024

  50. [51]

    Iacono, S

    M. Iacono, S. Weber, A. Glover, C. Bartolozzi, To- wards event-driven object detection with off-the- shelf deep learning, in: 2018 IEEE/RSJ Interna- tional Conference on Intelligent Robots and Sys- tems (IROS), IEEE, 2018, pp. 1–9

    work page 2018

  51. [52]

    YOLOv3: An Incremental Improvement

    J. Redmon, A. Farhadi, Yolov3: An incremen- tal improvement, arXiv preprint arXiv:1804.02767 (2018)

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  52. [53]

    M. Liu, M. Zhu, Mobile video object detection with temporally-aware feature maps, in: Proceed- ings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 5686–5695

    work page 2018

  53. [54]

    J. Li, J. Li, L. Zhu, X. Xiang, T. Huang, Y . Tian, Asynchronous spatio-temporal memory network for continuous event-based object detec- tion, IEEE Transactions on Image Processing 31 (2022) 2975–2987. 15

    work page 2022