pith. machine review for the scientific record. sign in

arxiv: 2604.23542 · v1 · submitted 2026-04-26 · 💻 cs.CV

Recognition: unknown

AusSmoke meets MultiNatSmoke: a fully-labelled diverse smoke segmentation dataset

Weihao Li , Hongjin Zhao , Gao Zhu , Ge-Peng Ji , Nicholas Wilson , Marta Yebra , Nick Barnes
Authors on Pith no claims yet

Pith reviewed 2026-05-08 06:47 UTC · model grok-4.3

classification 💻 cs.CV
keywords smoke segmentationwildfire detectiondatasetAusSmokeMultiNatSmokecomputer visiongeographic diversity
0
0 comments X

The pith

New AusSmoke dataset from Australia joins international data to create a ten-times-larger smoke segmentation benchmark.

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

The paper is trying to establish that collecting real smoke images from Australia and combining them with other international datasets creates a much larger and more diverse resource for training AI models to segment smoke in images. A sympathetic reader would care because existing datasets for this task are limited in size and geographic coverage, often relying on synthetic data, which makes it hard for models to work well in real-world wildfire situations around the world. The authors show through benchmarking that this new collection leads to better model performance and improved generalization to different locations. This matters for developing reliable early warning systems using cameras to detect wildfires quickly and reduce their impacts.

Core claim

We present AusSmoke, a new smoke segmentation dataset collected from Australia to address the data scarcity in this region. Furthermore, we introduce a MultiNational geographically diverse and substantially larger fully-labelled benchmark, called MultiNatSmoke, that consolidates publicly available international datasets with the newly collected Australian imagery, expanding the scale by an order of magnitude over previous collections. Finally, we benchmark smoke segmentation models, demonstrating improved performance and enhanced generalization across diverse geographical contexts.

What carries the argument

The integration of the new AusSmoke Australian smoke images with existing international datasets to form the larger MultiNatSmoke benchmark for training and evaluating smoke segmentation models.

If this is right

  • Smoke segmentation models achieve improved accuracy when trained on the expanded dataset.
  • Models exhibit better generalization to smoke appearances in varied geographical locations.
  • The use of real imagery reduces reliance on synthetic data for training.
  • The order-of-magnitude scale increase supports development of more robust detection systems.

Where Pith is reading between the lines

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

  • Dataset merging strategies like this may help in other areas of environmental AI where data is scarce.
  • Models could be tested in operational camera systems for real-time wildfire monitoring.
  • Additional validation datasets from new continents would strengthen the generalization evidence.

Load-bearing premise

The labels provided for the Australian images and the consolidated datasets are consistent and free of systematic biases that could affect model training or evaluation.

What would settle it

A test showing that models trained on MultiNatSmoke do not outperform those trained on previous datasets when evaluated on smoke images from a new, unseen region.

Figures

Figures reproduced from arXiv: 2604.23542 by Gao Zhu, Ge-Peng Ji, Hongjin Zhao, Marta Yebra, Nicholas Wilson, Nick Barnes, Weihao Li.

Figure 1
Figure 1. Figure 1: Example images on wildfire smoke segmentation from our AusSmoke dataset. view at source ↗
Figure 2
Figure 2. Figure 2: The cameras used for the AusSmoke dataset: on the view at source ↗
Figure 4
Figure 4. Figure 4: Region-wise distribution of the MultiNatSmoke. view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of smoke segmentation models at different training data scales. Larger training sets boost performance across view at source ↗
read the original abstract

Wildfires are an escalating global concern due to the devastating impacts on the environment, economy, and human health, with notable incidents such as the 2019-2020 Australian bushfires and the 2025 California wildfires underscoring the severity of these events. AI-enabled camera-based smoke detection has emerged as a promising approach for the rapid detection of wildfires. However, existing wildfire smoke segmentation datasets that are used for training detection and segmentation models are limited in scale, geographically constrained, and often rely on synthetic imagery, which hinders effective training and generalization. To overcome these limitations, we present AusSmoke, a new smoke segmentation dataset collected from Australia to address the data scarcity in this region. Furthermore, we introduce a MultiNational geographically diverse and substantially larger fully-labelled benchmark, called MultiNatSmoke, that consolidates publicly available international datasets with the newly collected Australian imagery, expanding the scale by an order of magnitude over previous collections. Finally, we benchmark smoke segmentation models, demonstrating improved performance and enhanced generalization across diverse geographical contexts. The project is available at \href{https://github.com/henryzhao0615/MultiNatSmoke}{Github}.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper introduces AusSmoke, a new fully-labelled smoke segmentation dataset collected from Australia to address regional data scarcity. It further presents MultiNatSmoke, a consolidated multinational benchmark that merges AusSmoke with existing international datasets, expanding the overall scale by an order of magnitude. The authors benchmark smoke segmentation models on these resources and claim improved performance together with enhanced generalization across diverse geographical contexts.

Significance. If the dataset labels prove consistent and the benchmarking protocols are shown to be uniform, the work would supply a substantially larger and geographically broader resource for training smoke segmentation models. This could meaningfully improve the robustness of AI-based early wildfire detection systems, particularly by mitigating the current limitations of small-scale or synthetic datasets and by filling gaps in underrepresented regions such as Australia.

major comments (2)
  1. Abstract: the headline claim of 'improved performance and enhanced generalization across diverse geographical contexts' is stated without any quantitative results, model architectures, baselines, training details, or performance tables, preventing verification of the central empirical contribution.
  2. MultiNatSmoke construction and benchmarking sections: the generalization claim requires demonstrated label consistency (no systematic annotation-style differences between AusSmoke and the consolidated international sets) and identical training protocols. No inter-annotator agreement scores, label-harmonization procedure, or ablation confirming fixed hyperparameters (augmentations, epochs, loss weighting) across sources are reported; apparent gains could therefore arise from data volume or protocol artifacts rather than geographical diversity.
minor comments (1)
  1. Abstract: the statement that MultiNatSmoke 'expands the scale by an order of magnitude over previous collections' would be strengthened by explicit numerical comparison of prior dataset sizes versus the new total.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed review. We address each major comment point by point below, indicating where revisions will be made to improve clarity and rigor.

read point-by-point responses

  1. Referee: Abstract: the headline claim of 'improved performance and enhanced generalization across diverse geographical contexts' is stated without any quantitative results, model architectures, baselines, training details, or performance tables, preventing verification of the central empirical contribution.

    Authors: We agree that the abstract would benefit from including quantitative context to support the headline claims. In the revised manuscript, we will update the abstract to briefly reference the segmentation models benchmarked, the consistent training protocols applied, and key performance metrics (such as mIoU gains on MultiNatSmoke relative to prior single-region datasets). This will allow immediate verification of the empirical contributions without altering the abstract's length or focus. revision: yes

  2. Referee: MultiNatSmoke construction and benchmarking sections: the generalization claim requires demonstrated label consistency (no systematic annotation-style differences between AusSmoke and the consolidated international sets) and identical training protocols. No inter-annotator agreement scores, label-harmonization procedure, or ablation confirming fixed hyperparameters (augmentations, epochs, loss weighting) across sources are reported; apparent gains could therefore arise from data volume or protocol artifacts rather than geographical diversity.

    Authors: We acknowledge the validity of this concern and the need for explicit evidence of label consistency and uniform protocols. The manuscript describes aligning AusSmoke annotations with the guidelines of the incorporated international datasets during MultiNatSmoke construction. However, we did not report inter-annotator agreement scores or provide hyperparameter ablations. In the revision, we will add a dedicated subsection on the label harmonization procedure, explicitly confirm that all benchmarking runs used identical hyperparameters, augmentations, epochs, and loss settings across sources (with a summary table), and include an ablation isolating the effect of geographical diversity from data volume. For inter-annotator agreement, we will report any available metrics from the new AusSmoke annotations and discuss consistency with prior datasets based on protocol alignment; where original annotations from external sources are unavailable for recomputation, we will note this limitation. revision: partial

Circularity Check

0 steps flagged

No circularity: dataset paper with empirical benchmarking only

full rationale

The paper presents a new dataset (AusSmoke) and a consolidated benchmark (MultiNatSmoke) followed by standard model benchmarking. No derivations, equations, predictions, fitted parameters, or first-principles claims appear in the abstract or described content. The work consists of data collection, consolidation of public datasets, and empirical evaluation of segmentation models. No load-bearing steps reduce to self-definition, fitted inputs renamed as predictions, or self-citation chains. The central claims rest on the existence and scale of the collected data plus observed benchmark numbers, which are externally verifiable through the released dataset and code rather than internally forced by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a dataset creation and benchmarking paper with no mathematical model, derivations, or theoretical claims, so the ledger is empty.

pith-pipeline@v0.9.0 · 5515 in / 1156 out tokens · 64136 ms · 2026-05-08T06:47:52.941161+00:00 · methodology

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

56 extracted references · 4 canonical work pages

  1. [1]

    The global fire atlas of individual fire size, duration, speed and di- rection.Earth System Science Data, 11(2):529–552,

    Niels Andela, Douglas C Morton, Louis Giglio, Ro- nan Paugam, Yang Chen, Stijn Hantson, Guido R Van Der Werf, and James T Randerson. The global fire atlas of individual fire size, duration, speed and di- rection.Earth System Science Data, 11(2):529–552,

  2. [2]

    Unprecedented burn area of australian mega forest fires.Nature Climate Change, 10(3):171– 172, 2020

    Matthias M Boer, Víctor Resco de Dios, and Ross A Bradstock. Unprecedented burn area of australian mega forest fires.Nature Climate Change, 10(3):171– 172, 2020. 2

    2020

  3. [3]

    Fire and smoke datasets in 20 years: An in-depth re- view.arXiv preprint arXiv:2503.14552, 2025

    Sayed Pedram Haeri Boroujeni, Niloufar Mehrabi, Fatemeh Afghah, Connor Peter McGrath, Danish Bhatkar, Mithilesh Anil Biradar, and Abolfazl Razi. Fire and smoke datasets in 20 years: An in-depth re- view.arXiv preprint arXiv:2503.14552, 2025. 3

    work page arXiv 2025

  4. [4]

    Cogent confabulation based expert system for seg- mentation and classification of natural landscape im- ages.Adv

    Maja Braovic, Darko Stipanicev, and Damir Krstinic. Cogent confabulation based expert system for seg- mentation and classification of natural landscape im- ages.Adv. Electr. Comput. Eng, 17(2):85–94, 2017. 4

    2017

  5. [5]

    Adaptive estimation of visual smoke detec- tion parameters based on spatial data and fire risk in- dex.Computer vision and image understanding, 118: 184–196, 2014

    Marin Bugari ´c, Toni Jakov ˇcevi´c, and Darko Sti- paniˇcev. Adaptive estimation of visual smoke detec- tion parameters based on spatial data and fire risk in- dex.Computer vision and image understanding, 118: 184–196, 2014. 2, 4

    2014

  6. [6]

    Channel and spatial attention based deep object co-segmentation

    Jia Chen, Yasong Chen, Weihao Li, Guoqin Ning, Mingwen Tong, and Adrian Hilton. Channel and spatial attention based deep object co-segmentation. Knowledge-Based Systems, 211:106550, 2021. 3

    2021

  7. [7]

    Encoder-decoder with atrous separable convolution for semantic image segmentation

    Liang-Chieh Chen, Yukun Zhu, George Papandreou, Florian Schroff, and Hartwig Adam. Encoder-decoder with atrous separable convolution for semantic image segmentation. InProceedings of the European con- ference on computer vision (ECCV), pages 801–818,

  8. [8]

    Masked- attention mask transformer for universal image seg- mentation

    Bowen Cheng, Ishan Misra, Alexander G Schwing, Alexander Kirillov, and Rohit Girdhar. Masked- attention mask transformer for universal image seg- mentation. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 1290–1299, 2022. 2, 3, 6

    2022

  9. [9]

    Bowfire: detection of fire in still images by integrating pixel color and tex- ture analysis

    Daniel YT Chino, Letricia PS Avalhais, Jose F Ro- drigues, and Agma JM Traina. Bowfire: detection of fire in still images by integrating pixel color and tex- ture analysis. In2015 28th SIBGRAPI conference on graphics, patterns and images, pages 95–102. IEEE,

  10. [10]

    An automatic fire detec- tion system based on deep convolutional neural net- works for low-power, resource-constrained devices

    Pedro Vinicius AB de Venancio, Adriano C Lisboa, and Adriano V Barbosa. An automatic fire detec- tion system based on deep convolutional neural net- works for low-power, resource-constrained devices. Neural Computing and Applications, 34(18):15349– 15368, 2022. 4, 5

    2022

  11. [11]

    Imagenet: A large-scale hierarchi- cal image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchi- cal image database. In2009 IEEE conference on com- puter vision and pattern recognition, pages 248–255. Ieee, 2009. 2, 6

    2009

  12. [12]

    Figlib & smokeynet: Dataset and deep learning model for real- time wildland fire smoke detection.Remote Sensing, 14(4):1007, 2022

    Anshuman Dewangan, Yash Pande, Hans-Werner Braun, Frank Vernon, Ismael Perez, Ilkay Altintas, Garrison W Cottrell, and Mai H Nguyen. Figlib & smokeynet: Dataset and deep learning model for real- time wildland fire smoke detection.Remote Sensing, 14(4):1007, 2022. 3, 4

    2022

  13. [13]

    Expert per- spective: Wildland fuels management would not have saved us from the january 2025 la fires, 2025

    Robert Fitch, Carla D’Antonio, Park Williams, Max Moritz, Shane Dewees, and Alex Hall. Expert per- spective: Wildland fuels management would not have saved us from the january 2025 la fires, 2025. Ac- cessed: 2025-05-16. 2

    2025

  14. [14]

    Preliminary results from a wildfire detection system using deep learning on re- mote camera images.Remote Sensing, 12(1):166,

    Kinshuk Govil, Morgan L Welch, J Timothy Ball, and Carlton R Pennypacker. Preliminary results from a wildfire detection system using deep learning on re- mote camera images.Remote Sensing, 12(1):166,

  15. [15]

    Mask r-cnn

    Kaiming He, Georgia Gkioxari, Piotr Dollár, and Ross Girshick. Mask r-cnn. InProceedings of the IEEE international conference on computer vision, pages 2961–2969, 2017. 3

    2017

  16. [16]

    Goss: Towards generalized open-set semantic seg- mentation.The Visual Computer, 40(4):2391–2404,

    Jie Hong, Weihao Li, Junlin Han, Jiyang Zheng, Pengfei Fang, Mehrtash Harandi, and Lars Petersson. Goss: Towards generalized open-set semantic seg- mentation.The Visual Computer, 40(4):2391–2404,

  17. [17]

    Segmentation models py- torch.https : / / github

    Pavel Iakubovskii. Segmentation models py- torch.https : / / github . com / qubvel / segmentation_models.pytorch, 2019. 6

    2019

  18. [18]

    Oneformer: One transformer to rule universal image segmenta- tion

    Jitesh Jain, Jiachen Li, Mang Tik Chiu, Ali Has- sani, Nikita Orlov, and Humphrey Shi. Oneformer: One transformer to rule universal image segmenta- tion. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 2989– 2998, 2023. 2, 6

    2023

  19. [19]

    Video polyp segmentation: A deep learning perspec- tive.Machine Intelligence Research, 19(6):531–549,

    Ge-Peng Ji, Guobao Xiao, Yu-Cheng Chou, Deng- Ping Fan, Kai Zhao, Geng Chen, and Luc Van Gool. Video polyp segmentation: A deep learning perspec- tive.Machine Intelligence Research, 19(6):531–549,

  20. [20]

    9 Deep gradient learning for efficient camouflaged ob- ject detection.Machine Intelligence Research, 20(1): 92–108, 2023

    Ge-Peng Ji, Deng-Ping Fan, Yu-Cheng Chou, Dengxin Dai, Alexander Liniger, and Luc Van Gool. 9 Deep gradient learning for efficient camouflaged ob- ject detection.Machine Intelligence Research, 20(1): 92–108, 2023. 3

    2023

  21. [21]

    Colon-x: Advancing intelligent colonoscopy from multimodal understanding to clinical reasoning

    Ge-Peng Ji, Jingyi Liu, Deng-Ping Fan, and Nick Barnes. Colon-x: Advancing intelligent colonoscopy from multimodal understanding to clinical reasoning. arXiv preprint arXiv:2512.03667, 2025. 2

    work page arXiv 2025

  22. [22]

    Frontiers in intelligent colonoscopy.Machine Intelli- gence Research, 2026

    Ge-Peng Ji, Jingyi Liu, Peng Xu, Nick Barnes, Fa- had Shahbaz Khan, Salman Khan, and Deng-Ping Fan. Frontiers in intelligent colonoscopy.Machine Intelli- gence Research, 2026. 2

    2026

  23. [23]

    Segment anything

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer Whitehead, Alexander C Berg, Wan-Yen Lo, et al. Segment anything. InProceedings of the IEEE/CVF international conference on computer vi- sion, pages 4015–4026, 2023. 5

    2023

  24. [24]

    Deep object co-segmentation

    Weihao Li, Omid Hosseini Jafari, and Carsten Rother. Deep object co-segmentation. InAsian Conference on Computer Vision, pages 638–653. Springer, 2018. 2, 3

    2018

  25. [25]

    Rein: Reusing imagenet to improve open-set object detection

    Weihao Li, Moshiur Farazi, Jie Hong, and Lars Pe- tersson. Rein: Reusing imagenet to improve open-set object detection. In2023 International Conference on Digital Image Computing: Techniques and Applica- tions (DICTA), pages 523–530. IEEE, 2023. 3

    2023

  26. [26]

    Microsoft coco: Com- mon objects in context

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Dollár, and C Lawrence Zitnick. Microsoft coco: Com- mon objects in context. InComputer vision–ECCV 2014: 13th European conference, zurich, Switzerland, September 6-12, 2014, proceedings, part v 13, pages 740–755. Springer, 2014. 2

    2014

  27. [27]

    Generalised co-salient object detection.arXiv preprint arXiv:2208.09668,

    Jiawei Liu, Jing Zhang, Ruikai Cui, Kaihao Zhang, Weihao Li, and Nick Barnes. Generalised co-salient object detection.arXiv preprint arXiv:2208.09668,

    work page arXiv

  28. [28]

    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. 6

    2021

  29. [29]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. InInternational Conference on Learning Representations, 2019. 6

    2019

  30. [30]

    Wildfire smoke dataset.https: //github.com/aiformankind/wildfire- smoke-dataset, 2020

    AI For Mankind. Wildfire smoke dataset.https: //github.com/aiformankind/wildfire- smoke-dataset, 2020. Accessed: 2025-05-15. 4

    2020

  31. [31]

    Understanding the black summer bushfires through research: A sum- mary of key findings from the bushfire and natural hazards crc, 2023

    Natural Hazards Research Australia. Understanding the black summer bushfires through research: A sum- mary of key findings from the bushfire and natural hazards crc, 2023. Accessed: 2025-05-16. 2

    2023

  32. [32]

    Bo- real forest fire: Uav-collected wildfire detection and smoke segmentation dataset.https : / / doi

    Julius Pesonen, Anna-Maria Raita-Hakola, Jukka Joutsalainen, Teemu Hakala, Waleed Akhtar, Väinö Karjalainen, Niko Koivumäki, Lauri Markelin, Juha Suomalainen, Raquel Alves de Oliveira, et al. Bo- real forest fire: Uav-collected wildfire detection and smoke segmentation dataset.https : / / doi . org/10.23729/fd-72c6cf74-b8eb-3687- 860d-bf93a1ab94c9, 2025. ...

    work page doi:10.23729/fd-72c6cf74-b8eb-3687- 2025

  33. [33]

    Firespot: A database for smoke detection in early-stage wildfires

    Natthaphol Pornpholkullapat, Warit Phankrawee, Per- aphat Boondet, Thin Lai Lai Thein, Phoummixay Si- harath, Jennifer Dela Cruz, Ken T Marata, Kanok- vate Tungpimolrut, and Jessada Karnjana. Firespot: A database for smoke detection in early-stage wildfires. In2023 18th International Joint Symposium on Arti- ficial Intelligence and Natural Language Proces...

    2023

  34. [34]

    Smokebench: Evaluating multimodal large language models for wildfire smoke detection

    Tianye Qi, Weihao Li, and Nick Barnes. Smokebench: Evaluating multimodal large language models for wildfire smoke detection. InProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2026. 3

    2026

  35. [35]

    A-M Raita-Hakola, S Rahkonen, J Suomalainen, L Markelin, R Oliveira, T Hakala, N Koivumäki, E Honkavaara, and I Pölönen. Combining yolo v5 and transfer learning for smoke-based wildfire detection in boreal forests.The International Archives of the Pho- togrammetry, Remote Sensing and Spatial Information Sciences, 48:1771–1778, 2023. 4

    2023

  36. [36]

    You only look once: Unified, real-time object detection

    Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi. You only look once: Unified, real-time object detection. InProceedings of the IEEE con- ference on computer vision and pattern recognition, pages 779–788, 2016. 3

    2016

  37. [37]

    Burned area semantic seg- mentation: A novel dataset and evaluation using con- volutional networks.ISPRS Journal of Photogramme- try and Remote Sensing, 202:565–580, 2023

    Tiago FR Ribeiro, Fernando Silva, José Moreira, and Rogério Luís de C Costa. Burned area semantic seg- mentation: A novel dataset and evaluation using con- volutional networks.ISPRS Journal of Photogramme- try and Remote Sensing, 202:565–580, 2023. 3

    2023

  38. [38]

    U-net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. InInternational Conference on Med- ical Image Computing and Computer-Assisted Inter- vention, pages 234–241. Springer, 2015. 2, 3, 6

    2015

  39. [39]

    Aerial imagery pile burn detection using deep learn- ing: The flame dataset.Computer Networks, 193: 108001, 2021

    Alireza Shamsoshoara, Fatemeh Afghah, Abolfazl Razi, Liming Zheng, Peter Z Fulé, and Erik Blasch. Aerial imagery pile burn detection using deep learn- ing: The flame dataset.Computer Networks, 193: 108001, 2021. 3

    2021

  40. [40]

    Gimo: Generative image outpainting for early smoke 10 segmentation

    Sahir Shrestha, Weihao Li, Gao Zhu, and Nick Barnes. Gimo: Generative image outpainting for early smoke 10 segmentation. InProceedings of the Synthetic Data for Computer Vision Workshop at CVPR, 2025. 3

    2025

  41. [41]

    SDI-Paste: Synthetic dynamic instance copy-paste

    Sahir Shrestha, Weihao Li, Gao Zhu, and Nick Barnes. SDI-Paste: Synthetic dynamic instance copy-paste. In Proceedings of the Synthetic Data for Computer Vi- sion Workshop at CVPR, 2025. 3

    2025

  42. [42]

    Computer vision for wildfire research: An evolving image dataset for processing and analysis.Fire Safety Journal, 92:188– 194, 2017

    Tom Toulouse, Lucile Rossi, Antoine Campana, Tur- gay Celik, and Moulay A Akhloufi. Computer vision for wildfire research: An evolving image dataset for processing and analysis.Fire Safety Journal, 92:188– 194, 2017. 3

    2017

  43. [43]

    High-severity wildfires in temperate australian forests have increased in extent and aggregation in recent decades.PloS one, 15(11): e0242484, 2020

    Bang Nguyen Tran, Mihai A Tanase, Lauren T Ben- nett, and Cristina Aponte. High-severity wildfires in temperate australian forests have increased in extent and aggregation in recent decades.PloS one, 15(11): e0242484, 2020. 2

    2020

  44. [44]

    Ung dung tri tue nhan tao de phat hien bat thuong trong giam sat rung.Journal of Science and Technology on Information and Communications, 1(4):118–124,

    Quang Vinh Vu, Cong Tran, and Dat Anh Tran. Ung dung tri tue nhan tao de phat hien bat thuong trong giam sat rung.Journal of Science and Technology on Information and Communications, 1(4):118–124,

  45. [45]

    Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pier- ric Cistac, Tim Rault, Rémi Louf, Morgan Funtow- icz, Joe Davison, Sam Shleifer, Patrick von Platen, Clara Ma, Yacine Jernite, Julien Plu, Canwen Xu, Teven Le Scao, Sylvain Gugger, Mariama Drame, Quentin Lhoest, and Alexander M. Rush. Trans- formers: State-of-t...

    2020

  46. [46]

    Segformer: Simple and efficient design for semantic segmentation with transformers.Advances in neural information processing systems, 34:12077–12090, 2021

    Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anand- kumar, Jose M Alvarez, and Ping Luo. Segformer: Simple and efficient design for semantic segmentation with transformers.Advances in neural information processing systems, 34:12077–12090, 2021. 2, 3, 6

    2021

  47. [47]

    Self-training with noisy student im- proves imagenet classification

    Qizhe Xie, Minh-Thang Luong, Eduard Hovy, and Quoc V Le. Self-training with noisy student im- proves imagenet classification. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10687–10698, 2020. 5

    2020

  48. [48]

    Transmission-guided bayesian generative model for smoke segmentation

    Siyuan Yan, Jing Zhang, and Nick Barnes. Transmission-guided bayesian generative model for smoke segmentation. InProceedings of the AAAI Conference on Artificial Intelligence, pages 3009–3017, 2022. 2, 3, 4, 5, 6

    2022

  49. [49]

    Fosp: Focus and separation network for early smoke segmentation

    Lujian Yao, Haitao Zhao, Jingchao Peng, Zhongze Wang, and Kaijie Zhao. Fosp: Focus and separation network for early smoke segmentation. InProceed- ings of the AAAI Conference on Artificial Intelligence, pages 6621–6629, 2024. 2, 3, 4, 5, 6

    2024

  50. [50]

    Technological solutions for living with fire in the age of megafires.One Earth, 7(6):932–935, 2024

    Marta Yebra, Robert Mahony, and Robert Debus. Technological solutions for living with fire in the age of megafires.One Earth, 7(6):932–935, 2024. 2

    2024

  51. [51]

    Deep smoke seg- mentation.Neurocomputing, 357:248–260, 2019

    Feiniu Yuan, Lin Zhang, Xue Xia, Boyang Wan, Qinghua Huang, and Xuelong Li. Deep smoke seg- mentation.Neurocomputing, 357:248–260, 2019. 2, 3

    2019

  52. [52]

    Wildland forest fire smoke de- tection based on faster r-cnn using synthetic smoke images.Procedia engineering, 211:441–446, 2018

    Qi-xing Zhang, Gao-hua Lin, Yong-ming Zhang, Gao Xu, and Jin-jun Wang. Wildland forest fire smoke de- tection based on faster r-cnn using synthetic smoke images.Procedia engineering, 211:441–446, 2018. 4

    2018

  53. [53]

    False alarm rectification for early smoke seg- mentation

    Hongjin Zhao, Weihao Li, Ge-Peng Ji, and Nick Barnes. False alarm rectification for early smoke seg- mentation. InProceedings of the IEEE/CVF Win- ter Conference on Applications of Computer Vision (WACV), 2026. 3

    2026

  54. [54]

    Der- meval: A dermatologist-reviewed benchmark for mul- timodal large language models

    Hongjin Zhao, Weihao Li, Zhenyue Qin, Ge-Peng Ji, Yang Liu, Tom Gedeon, and Nick Barnes. Der- meval: A dermatologist-reviewed benchmark for mul- timodal large language models. InProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2026. 2

    2026

  55. [55]

    Towards open-set object detection and discovery

    Jiyang Zheng, Weihao Li, Jie Hong, Lars Petersson, and Nick Barnes. Towards open-set object detection and discovery. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 3961–3970, 2022. 3

    2022

  56. [56]

    Scene parsing through ade20k dataset

    Bolei Zhou, Hang Zhao, Xavier Puig, Sanja Fidler, Adela Barriuso, and Antonio Torralba. Scene parsing through ade20k dataset. InProceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 633–641, 2017. 6 11

    2017