Pith. sign in

REVIEW 3 major objections 3 minor 45 references

Reliable Smoke Detection via Optical Flow-Guided Feature Fusion and Transformer-Based Uncertainty Modeling

T0 review · 3 major / 3 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read A single-camera pipeline claims to segment smoke from video and flag, with a calibrated confidence score, when its own prediction may be wrong.

desk verdict The smoke-detection accuracy and calibration claims are built on a self-referential evaluation loop, so the headline numbers are not trustworthy; the optical flow module is externally validated, but that is not the paper's central claim. read the letter →

arxiv 2508.14597 v1 pith:BLGYKVZR submitted 2025-08-20 cs.CV

Nitish Kumar Mahala , Muzammil Khan , Pushpendra Kumar This is my paper · ORCID
classification cs.CV
keywords smokedetectionopticalflowlevel-setvariationalmodelSwintransformeruncertaintyestimationGaussianmixturesegmentationearlyfiremonocularvideo
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that smoke can be detected reliably from ordinary single-camera video, without multi-sensor hardware, by fusing motion and appearance cues and letting a transformer both segment the smoke and measure its own uncertainty. The training labels are generated automatically: a fractional-order variational optical flow model color-encodes motion while preserving smoke-boundary discontinuities, and a Gaussian Mixture Model binarizes those maps into smoke masks. A two-phase shifted-windows transformer first learns accurate segmentation, then learns to output aleatoric and epistemic uncertainty, with the claim that the resulting confidence is well-calibrated, tracks error, and reflects per-class difficulty. If true, surveillance and industrial safety systems could issue early fire warnings with an explicit reliability signal attached to every prediction.

What carries the argument

The load-bearing object is the two-phase uncertainty-aware shifted-windows transformer (TP-UAST), built on a Swin Transformer backbone with a multi-scale uncertainty head. Its labels come from two coupled mechanisms: an optical-flow stage (FCDLe-FOV, a four-color-theorem-inspired dual-phase level-set fractional-order variational model) that color-encodes motion while preserving discontinuities, and a Gaussian Mixture Model that fuses the flow color maps with appearance cues into binary smoke masks. The two-phase training regimen is the third mechanism: phase one fixes segmentation accuracy, and phase two fits aleatoric and epistemic uncertainty so the model outputs a plausibility score along

What would settle it

Run the published flow-to-GMM-to-transformer pipeline on a human-annotated test set of real fire and smoke videos and compare per-pixel agreement against the manual masks; if agreement falls well short of the reported numbers, or if the model outputs high-confidence smoke masks on smoke-free clips containing moving vehicles or pedestrians, the synthetic-label premise is broken. A simpler version: a video of a car crossing a smoke-free scene should produce no smoke mask, so any substantial mask from non-smoke motion directly indicts the GMM stage.

Watch

Extended reading notes

Core claim

The central claim is that a Two-Phase Uncertainty-Aware Shifted Windows Transformer (TP-UAST), trained on masks produced by its own optical-flow-plus-GMM pipeline, both segments smoke and estimates how much its own prediction should be trusted. The optical flow stage — a four-color-theorem-inspired dual-phase level-set fractional-order variational model — encodes motion into color maps while preserving the sharp discontinuities at smoke boundaries; a Gaussian Mixture Model then fuses these motion maps with appearance cues into binary smoke masks. The first training phase optimizes segmentation accuracy; the second teaches the model to emit a plausibility confidence per prediction by jointly

Load-bearing premise

The training labels and the evaluation ground truth are both produced by the paper's own pipeline — a Gaussian Mixture Model binarizing its own optical-flow color maps — so the reported accuracy and confidence calibration presuppose that these synthetic masks faithfully identify smoke and nothing else.

Editorial extensions

If this is right

  • Early-warning alarm systems could run on ordinary surveillance footage from a single camera, with no multi-sensor array, because smoke precursors are detected from motion-encoded monocular video.
  • Every prediction carries a plausibility confidence, so alarm logic can defer to human review precisely when the model reports high uncertainty rather than firing on all detections.
  • The smoke-segmentation dataset is produced automatically from flow and color cues, removing the need for costly manual pixel-level annotation of smoke.
  • The flow model's discontinuity preservation is claimed to yield sharper smoke boundaries than prior variational and learned optical-flow baselines on Middlebury and Sintel benchmarks.
  • Separating accuracy from confidence lets the calibration phase be re-run or extended without retraining the segmentation head, since the two tasks are optimized in distinct phases.

Reading between the lines

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

  • The paper leaves open whether its calibration generalizes: because both training and evaluation use the same synthetic masks, 'well-calibrated' may mean consistent with the pipeline's own labels rather than with human judgment of smoke; a human-annotated held-out set is the decisive check.
  • The identical flow-to-GMM labeling trick should transfer to other semi-transparent moving phenomena such as steam, fog, or dust, where manual segmentation is equally hard — with the same circular-label caveat applying.
  • Because the uncertainty head separates aleatoric from epistemic components, the framework is a natural fit for active learning: the model could request human labels for the frames where epistemic uncertainty is highest.
  • Deployment risk concentrates in the motion prior: if the GMM stage labels any moving object (vehicles, trees, pedestrians) as smoke, the reported accuracy measures agreement with that prior, not true smoke detection, so domain shift could silently break the system.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 3 minor

Summary. The paper proposes TP-UAST, a Two-Phase Uncertainty-Aware Shifted Windows Transformer for smoke segmentation from monocular video. The pipeline first estimates optical flow with a fractional-order variational model, color-encodes the flow, fuses it with appearance cues via a Gaussian Mixture Model (GMM) to produce binary smoke masks, and then trains a Swin-Transformer-based segmentation network with a two-phase loss: first for segmentation accuracy, then for aleatoric and epistemic uncertainty estimation. The authors claim superior generalization, robustness, and well-calibrated uncertainty relative to state-of-the-art smoke detectors. The only external validation reported is for the optical-flow component on Middlebury and Sintel; the smoke segmentation and calibration claims are evaluated using masks generated by the same GMM pipeline that is part of the proposed system.

Significance. If the central evaluation were independent, the paper would contribute a useful integration of optical-flow-guided pseudo-labeling, transformer-based segmentation, and joint aleatoric/epistemic uncertainty estimation for early fire detection. The optical flow module is benchmarked against external datasets (Table 1), and the two-phase uncertainty training is a sensible architectural choice. However, the significance of the paper hinges on the validity of the smoke masks used as ground truth. Because those masks are generated by the paper's own front-end, the reported segmentation accuracy and calibration metrics currently measure how well the transformer reproduces the GMM labels, not how well it detects actual smoke. Independent validation is therefore essential before the claimed reliability can be accepted.

major comments (3)
  1. [Abstract and p. 13, 'Fourth experiment'] The smoke masks used to train and evaluate TP-UAST are produced by the same pipeline that is being proposed. The abstract states that 'color-encoded optical flow maps are fused with appearance cues via a Gaussian Mixture Model to generate binary segmentation masks of the smoke regions,' and p. 13 confirms: 'This binary mask is derived from the estimated color maps using the GMM technique.' These masks serve as ground truth for both training and scoring the transformer, making the reported accuracy and calibration self-referential. If the front-end systematically labels non-smoke motion (vehicles, trees, pedestrians) as smoke, the high scores and 'well-calibrated' uncertainty are artifacts of the front-end bias rather than evidence of reliable smoke detection. Please evaluate on human-annotated smoke segmentation datasets, or at least on a manually validated subset, and report per-image a
  2. [Table 1] The only external benchmarks reported (Middlebury, Sintel) validate the optical flow model, not the smoke segmentation or the uncertainty calibration. Table 1 therefore does not support the abstract's claim of 'superior generalization and robustness' for the detection system as a whole. A quantitative comparison against established smoke/fire detection benchmarks or human-annotated smoke masks is needed to substantiate the central claim. Without such independent evaluation, the reported detection performance cannot be distinguished from the behavior of the GMM labeling step.
  3. [pp. 15–17, Figs. 11–12] The claim that TP-UAST's 'uncertainty estimates are well-calibrated, informative of error, and sensitive to class-dependent difficulty' is made using the same self-generated labels discussed above. In addition, no quantitative calibration metric (e.g., Expected Calibration Error, Brier score) is reported; the evidence consists of reliability diagrams and histograms. Please report standard calibration metrics and compare them with baseline uncertainty methods, using independent ground-truth labels rather than the pipeline's own masks.
minor comments (3)
  1. [Dataset description] The 'novel smoke segmentation dataset' is mentioned in the abstract but not described in sufficient detail. Please include image/video counts, sources, resolution, split information, and the rationale for the GMM component count and binarization threshold. Without this information, the experiments are not reproducible and the sensitivity of the generated labels to these choices cannot be assessed.
  2. [Figures 7–8] Qualitative comparisons on smoke and non-smoke data are presented, but no quantitative false-positive / false-negative rates are given for the non-smoke cases. A quantitative analysis of false positives on non-smoke videos would strengthen the reliability claim.
  3. [General presentation] Several passages of the manuscript text are garbled or contain unreadable characters, making it difficult to verify the mathematical formulation in Sections 2–3. A clean, correctly rendered version should be provided.

Circularity Check

1 steps flagged · score 7.0 of 10

Evaluation loop: smoke ground-truth masks are generated by the paper's own GMM/optical-flow pipeline, so TP-UAST's accuracy and calibration claims measure self-reproduction rather than external smoke detection.

  1. self definitional [Abstract; Section 4 (fourth experiment), p. 13; Section 5, p. 15]
    "The resulting color-encoded optical flow maps are fused with appearance cues via a Gaussian Mixture Model to generate binary segmentation masks of the smoke regions. ... This binary mask is derived from the estimated color maps using the GMM technique, which effectively distinguishes between smoke and non-smoke regions based ... Together, these analyses demonstrate that TP-UAST’s uncertainty estimates are well-calibrated, informative of error, and sensitive to class-dependent difficulty."

    The binary masks used as ground truth for both training and evaluation are not independent labels; they are outputs of the paper's own front-end: the GMM binarizes the authors' color-encoded optical flow maps. TP-UAST is therefore trained and scored against masks generated by the same pipeline it is supposed to validate. Reported segmentation accuracy and uncertainty calibration quantify how well TP-UAST reproduces the GMM's own labels, not how well it detects real smoke. The only external validation, Table 1 on Middlebury and Sintel, tests the optical flow model, not the smoke masks. If the GMM systematically mislabels non-smoke motion as smoke, the high metrics and 'well-calibrated' uncertainty are artifacts of the label generator, not evidence of reliable smoke detection.

full rationale

The central circularity is the smoke dataset construction: the paper's abstract states that GMM fusion generates binary smoke masks, and page 13 confirms these masks are derived from the estimated color maps using the GMM technique. Those same masks are used as ground truth for training and for evaluating TP-UAST's segmentation and uncertainty. This makes the reliability claim self-referential: the model is being asked to predict labels produced by the same optical-flow/GMM front-end, and the metrics measure agreement with that front-end rather than with independent smoke annotations. The optical-flow component is self-cited ([5], [24]), but it receives external validation on Middlebury and Sintel in Table 1, so that self-citation is not itself load-bearing. The paper does compare against other methods on its own dataset, so relative rankings are not vacuous; however, the absolute accuracy, robustness, and calibration claims are undermined by the lack of externally grounded ground truth. No other circular steps were found: the transformer architecture follows standard Swin/attention designs, and the uncertainty formulation is tested only against the self-generated labels. Score 7 reflects a central evaluation loop that is partial rather than total: the transformer could still fail against these labels, but the target labels are by construction outputs of the paper's own pipeline.

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

The paper's main intellectual debts are to its own prior optical flow work ([5], [9], [24]) and to standard components (GMM [7], Swin [30], level sets [28]). No new physical entities are postulated. The central free parameter is effectively the entire label-generation pipeline, since the GMM hyperparameters and flow regularizer determine the ground truth the transformer is judged against.

free parameters (3)
  • GMM component count and binarization threshold
    The Gaussian Mixture Model converts color-encoded flow plus appearance features into binary smoke masks (abstract; p. 13). Count and threshold are not stated in the readable text but determine the ground truth itself.
  • Optical flow regularization weights and fractional order
    The dual-phase level-set fractional-order variational flow (FCDLe-FOV) carries regularization and fractional-order hyperparameters, tuned against Middlebury/Sintel benchmarks in Table 1; these inherit from self-cited [5], [24].
  • Two-phase training loss weights
    Phase 1 optimizes segmentation accuracy, phase 2 adds aleatoric and epistemic uncertainty losses (abstract); the weighting between these terms is a free choice that shapes the calibration claims.
assumptions (5)
  • standard math Vese-Chan multiphase level set formulation correctly represents disjoint motion regions
    The dual-phase level-set optical flow model builds on [28]; invoked for FCDLe-FOV.
  • standard math Chambolle projection converges for the fractional-order variational functional
    Solver relied on per [6], visible only via the reference list; convergence for the fractional-order variant is asserted.
  • domain assumption GMM over color-encoded optical flow separates smoke from non-smoke motion
    This is the load-bearing premise behind dataset construction: it assumes smoke occupies a distinguishable cluster in flow-plus-appearance feature space, and this premise defines the ground truth.
  • domain assumption Optical flow motion statistics correlate with smoke plumes
    The whole motion-encoding stage presumes smoke has distinctive kinematics (abstract).
  • domain assumption Swin transformer inductive bias transfers to smoke segmentation
    The architecture [30] is adopted without re-derivation for the smoke domain.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Reliable Smoke Detection via Optical Flow-Guided Feature Fusion and Transformer-Based Uncertainty Modeling." pith.science (2026). https://pith.science/paper/BLGYKVZR

@misc{pith2026250814597,
  author       = {Pith},
  title        = {Pith review of: Reliable Smoke Detection via Optical Flow-Guided Feature Fusion and Transformer-Based Uncertainty Modeling},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BLGYKVZR}},
  note         = {Machine review of arXiv:2508.14597}
}
read the original abstract

Fire outbreaks pose critical threats to human life and infrastructure, necessitating high-fidelity early-warning systems that detect combustion precursors such as smoke. However, smoke plumes exhibit complex spatiotemporal dynamics influenced by illumination variability, flow kinematics, and environmental noise, undermining the reliability of traditional detectors. To address these challenges without the logistical complexity of multi-sensor arrays, we propose an information-fusion framework by integrating smoke feature representations extracted from monocular imagery. Specifically, a Two-Phase Uncertainty-Aware Shifted Windows Transformer for robust and reliable smoke detection, leveraging a novel smoke segmentation dataset, constructed via optical flow-based motion encoding, is proposed. The optical flow estimation is performed with a four-color-theorem-inspired dual-phase level-set fractional-order variational model, which preserves motion discontinuities. The resulting color-encoded optical flow maps are fused with appearance cues via a Gaussian Mixture Model to generate binary segmentation masks of the smoke regions. These fused representations are fed into the novel Shifted-Windows Transformer, which is augmented with a multi-scale uncertainty estimation head and trained under a two-phase learning regimen. First learning phase optimizes smoke detection accuracy, while during the second phase, the model learns to estimate plausibility confidence in its predictions by jointly modeling aleatoric and epistemic uncertainties. Extensive experiments using multiple evaluation metrics and comparative analysis with state-of-the-art approaches demonstrate superior generalization and robustness, offering a reliable solution for early fire detection in surveillance, industrial safety, and autonomous monitoring applications.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 41 canonical work pages

  1. [1]

    Deployment and integration of smart sensors with iot devices detecting fire disasters in huge forest environment

    Fengmei Cui. Deployment and integration of smart sensors with iot devices detecting fire disasters in huge forest environment. Computer Communications, 150:818–827, 2020

    work page 2020

  2. [2]

    A new multi-sensor fire detection method based on lstm networks with environmental information fusion

    Pingshan Liu, Pingchuan Xiang, and Dianjie Lu. A new multi-sensor fire detection method based on lstm networks with environmental information fusion. Neural Computing and Applications, 35(36):25275–25289, 2023

    work page 2023

  3. [3]

    From smoke to fire: A forest fire early warning and risk assessment model fusing multimodal data

    Peixian Jin, Pengle Cheng, Xiaodong Liu, and Ying Huang. From smoke to fire: A forest fire early warning and risk assessment model fusing multimodal data. Engineering Applications of Artificial Intelligence, 152:110848, 2025

    work page 2025

  4. [4]

    Optical flow with fractional order regularization: variational model and solution method

    Somayeh Gh Bardeji, Isabel N Figueiredo, and Ercília Sousa. Optical flow with fractional order regularization: variational model and solution method. Applied Numerical Mathematics, 114:188–200, 2017

    work page 2017

  5. [5]

    A level set based fractional order variational model for motion estimation in application oriented spectrum

    Khan Muzammil and Kumar Pushpendra. A level set based fractional order variational model for motion estimation in application oriented spectrum. Expert Systems with Applications, 219:119628, 2023

    work page 2023

  6. [6]

    An algorithm for total variation minimization and applications

    Antonin Chambolle. An algorithm for total variation minimization and applications. Journal of Mathematical Imaging and Vision, 20:89–97, 2004

    work page 2004

  7. [7]

    Image segmentation using gaussian mixture model

    R Farnoush and PAK B ZAR. Image segmentation using gaussian mixture model. 2008

    work page 2008

  8. [8]

    Improving Uncertainty-Error Correspondence in Deep Bayesian Medical Image Segmentation

    Prerak Mody, Nicolas F Chaves-de Plaza, Chinmay Rao, Eleftheria Astrenidou, Mischa de Ridder, Nienke Hoekstra, Klaus Hildebrandt, and Marius Staring. Improving uncertainty-error correspondence in deep bayesian medical image segmentation. arXiv preprint arXiv:2409.03470, 2024

    work page Pith review arXiv 2024

Show all 45 references
  1. [9]

    Cnn-based fire prediction using fractional order optical flow and smoke features

    Muzammil Khan, Pushpendra Kumar, and Nitish Kumar Mahala. Cnn-based fire prediction using fractional order optical flow and smoke features. In Applications of Optimization and Machine Learning in Image Processing and IoT, pages 156–180. Chapman and Hall/CRC, 2023

    2023

  2. [10]

    A deep separable convolutional neural network for multiscale image-based smoke detection

    Yinuo Huo, Qixing Zhang, Yang Jia, Dongcai Liu, Jinfu Guan, Gaohua Lin, and Yongming Zhang. A deep separable convolutional neural network for multiscale image-based smoke detection. Fire Technology, pages 1–24, 2022

    2022

  3. [11]

    An image processing technique for automatically detecting forest fire

    Jerome Vicente and Philippe Guillemant. An image processing technique for automatically detecting forest fire. International Journal of Thermal Sciences, 41(12):1113–1120, 2002

    2002

  4. [12]

    Deep dual-channel neural network for image-based smoke detection

    Ke Gu, Zhifang Xia, Junfei Qiao, and Weisi Lin. Deep dual-channel neural network for image-based smoke detection. IEEE Transactions on Multimedia, 22(2):311–323, 2019

    2019

  5. [13]

    Image fire detection algorithms based on convolutional neural networks

    Pu Li and Wangda Zhao. Image fire detection algorithms based on convolutional neural networks. Case Studies in Thermal Engineering, 19:100625, 2020

    2020

  6. [14]

    Smoke detection on video sequences using 3d convolutional neural networks

    Gaohua Lin, Yongming Zhang, Gao Xu, and Qixing Zhang. Smoke detection on video sequences using 3d convolutional neural networks. Fire Technology, 55:1827–1847, 2019

    2019

  7. [15]

    Convolution-enhanced vision transformer network for smoke recognition

    Guangtao Cheng, Yancong Zhou, Shan Gao, Yingyu Li, and Hao Yu. Convolution-enhanced vision transformer network for smoke recognition. Fire Technology, 59(2):925–948, 2023

    2023

  8. [16]

    Video smoke detection method based on cell root–branch structure

    Huajun Song and Yulin Chen. Video smoke detection method based on cell root–branch structure. Signal, Image and Video Processing, pages 1–9, 2024

    2024

  9. [17]

    Fire and smoke detection in complex environments

    Furkat Safarov, Shakhnoza Muksimova, Misirov Kamoliddin, and Young Im Cho. Fire and smoke detection in complex environments. Fire, 7(11):389, 2024

    2024

  10. [18]

    Transformer-based fire detection in videos

    Konstantina Mardani, Nicholas Vretos, and Petros Daras. Transformer-based fire detection in videos. Sensors, 23(6):3035, 2023

    2023

  11. [19]

    Optical flow estimation for flame detection in videos

    Martin Mueller, Peter Karasev, Ivan Kolesov, and Allen Tannenbaum. Optical flow estimation for flame detection in videos. IEEE Transactions on Image Processing, 22(7):2786–2797, 2013

    2013

  12. [20]

    Deep belief network for smoke detection

    Arun Singh Pundir and Balasubramanian Raman. Deep belief network for smoke detection. Fire technology, 53:1943–1960, 2017

    1943

  13. [21]

    A modified variational functional for estimating dense and discontinuity preserving optical flow in various spectrum

    Pushpendra Kumar and Sanjeev Kumar. A modified variational functional for estimating dense and discontinuity preserving optical flow in various spectrum. AEU-International Journal of Electronics and Communications, 70(3):289–300, 2016

    2016

  14. [22]

    Non-local weighted regularization for optical flow estimation.Optik, 208:164069, 2020

    Zhenghua Huang and Aimin Pan. Non-local weighted regularization for optical flow estimation.Optik, 208:164069, 2020. 20 A PREPRINT - S EPTEMBER 12, 2025

    2020

  15. [23]

    Arnisha Khondaker, Arman Khandaker, and Jia Uddin. Computer vision-based early fire detection using enhanced chromatic segmentation and optical flow analysis technique.International Arab Journal of Information Technology, 17(6):947–953, 2020

    2020

  16. [24]

    Caputo derivative based nonlinear fractional order variational model for motion estimation in various application oriented spectrum

    Muzammil Khan, Nitish Kumar Mahala, and Pushpendra Kumar. Caputo derivative based nonlinear fractional order variational model for motion estimation in various application oriented spectrum. S¯adhan¯a, 49(1):1–28, 2024

    2024

  17. [25]

    Video fire smoke detection using motion and color features

    Yu Chunyu, Fang Jun, Wang Jinjun, and Zhang Yongming. Video fire smoke detection using motion and color features. Fire technology, 46:651–663, 2010

    2010

  18. [26]

    Video smoke detection base on dense optical flow and convolutional neural network

    Yuanlu Wu, Minghao Chen, Yan Wo, and Guoqiang Han. Video smoke detection base on dense optical flow and convolutional neural network. Multimedia Tools and Applications, 80:35887–35901, 2021

    2021

  19. [27]

    A daytime smoke detection method based on variances of optical flow and characteristics of hsv color on footage from outdoor camera in urban city

    Kazutaka Kikuta, Ken T Murata, and Yuki Murakami. A daytime smoke detection method based on variances of optical flow and characteristics of hsv color on footage from outdoor camera in urban city. Fire Technology, 60(3):1427–1452, 2024

    2024

  20. [28]

    A multiphase level set framework for image segmentation using the mumford and shah model

    Luminita A Vese and Tony F Chan. A multiphase level set framework for image segmentation using the mumford and shah model. International journal of computer vision, 50:271–293, 2002

    2002

  21. [29]

    V on neumann stability analysis for multi-level multi-step methods

    Arun Govind Neelan. V on neumann stability analysis for multi-level multi-step methods. arXiv preprint arXiv:2310.08274, 2023

    2023 arXiv

  22. [30]

    Swin transformer: Hierarchical vision transformer using shifted windows

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. InProceedings of the IEEE/CVF international conference on computer vision, pages 10012–10022, 2021

    2021

  23. [31]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need. corr abs/1706.03762, 2017

    2017 arXiv

  24. [32]

    Vit-saps: Detail-aware transformer for mechanical assembly semantic segmentation

    Haitao Dong, Chengjun Chen, Jinlei Wang, Feixiang Shen, and Yong Pang. Vit-saps: Detail-aware transformer for mechanical assembly semantic segmentation. IEEE Access, 11:41467–41479, 2023

    2023

  25. [33]

    2d positional embedding-based transformer for scene text recognition

    Zobeir Raisi, Mohamed A Naiel, Paul Fieguth, Steven Wardell, and John Zelek. 2d positional embedding-based transformer for scene text recognition. Journal of Computational Vision and Imaging Systems, 6(1):1–4, 2020

    2020

  26. [34]

    A field-segmentation-based variational optical flow method for piv measurements of nonuniform flows

    Jin Lu, Hua Yang, Qinghu Zhang, and Zhouping Yin. A field-segmentation-based variational optical flow method for piv measurements of nonuniform flows. Experiments in Fluids, 60:1–17, 2019

    2019

  27. [35]

    Tfnet: Transformer- based multi-scale feature fusion forest fire image detection network

    Hongying Liu, Fuquan Zhang, Yiqing Xu, Junling Wang, Hong Lu, Wei Wei, and Jun Zhu. Tfnet: Transformer- based multi-scale feature fusion forest fire image detection network. Fire, 8(2):59, 2025

    2025

  28. [36]

    Efficient detection of forest fire smoke in uav aerial imagery based on an improved yolov5 model and transfer learning

    Huanyu Yang, Jun Wang, and Jiacun Wang. Efficient detection of forest fire smoke in uav aerial imagery based on an improved yolov5 model and transfer learning. Remote Sensing, 15(23):5527, 2023

    2023

  29. [37]

    Edgefiresmoke: A novel lightweight cnn model for real-time video fire–smoke detection

    Jefferson Silva Almeida, Chenxi Huang, Fabrício Gonzalez Nogueira, Surbhi Bhatia, and Victor Hugo C de Al- buquerque. Edgefiresmoke: A novel lightweight cnn model for real-time video fire–smoke detection. IEEE Transactions on Industrial Informatics, 18(11):7889–7898, 2022

    2022

  30. [38]

    Real-time video fire/smoke detection based on cnn in antifire surveillance systems

    Sergio Saponara, Abdussalam Elhanashi, and Alessio Gagliardi. Real-time video fire/smoke detection based on cnn in antifire surveillance systems. Journal of Real-Time Image Processing, 18:889–900, 2021

    2021

  31. [39]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In IEEE conference on computer vision and pattern recognition, pages 770–778, 2016

    2016

  32. [40]

    Very deep convolutional networks for large-scale image recognition

    Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556, 2014

    2014 arXiv

  33. [41]

    Multi-level feature fusion networks for smoke recognition in remote sensing imagery

    Yupeng Wang, Yongli Wang, Zaki Ahmad Khan, Anqi Huang, and Jianghui Sang. Multi-level feature fusion networks for smoke recognition in remote sensing imagery. Neural Networks, 184:107112, 2025

    2025

  34. [42]

    Fire and smoke real-time detection algorithm for coal mines based on improved yolov8s

    Derui Kong, Yinfeng Li, and Manzhen Duan. Fire and smoke real-time detection algorithm for coal mines based on improved yolov8s. Plos one, 19(4):e0300502, 2024

    2024

  35. [43]

    Indoor fire and smoke detection based on optimized yolov5

    Md Shafak Shahriar Sozol, M Rubaiyat Hossain Mondal, and Achmad Husni Thamrin. Indoor fire and smoke detection based on optimized yolov5. PLoS One, 20(4):e0322052, 2025

    2025

  36. [44]

    Mobilenets: Efficient convolutional neural networks for mobile vision applications

    Andrew G Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco Andreetto, and Hartwig Adam. Mobilenets: Efficient convolutional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861, 2017

    2017 arXiv

  37. [45]

    On the performance of googlenet and alexnet applied to sketches

    Pedro Ballester and Ricardo Araujo. On the performance of googlenet and alexnet applied to sketches. In Proceedings of the AAAI conference on artificial intelligence, volume 30, 2016. 21

    2016

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.