Pith. sign in

REVIEW 5 major objections 6 minor 27 references

Seeing Heat with Color -- RGB-Only Wildfire Temperature Inference from SAM-Guided Multimodal Distillation using Radiometric Ground Truth

T0 review · 5 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read RGB-only drone images can predict wildfire pixel temperatures after multimodal distillation.

desk verdict First RGB-only wildfire temperature regression task, but the evaluation uses the same auto-masks for both training and scoring, so the headline numbers are not yet trustworthy. read the letter →

arxiv 2505.01638 v1 pith:WDPLL3KS submitted 2025-05-03 eess.IV cs.AIcs.CV

Michael Marinaccio , Fatemeh Afghah This is my paper · ORCID
classification eess.IVcs.AIcs.CV
keywords wildfiretemperaturepredictionRGB-onlyinferenceknowledgedistillationradiometricthermalTIFFSAM-guidedpseudo-labelsTOPSISmaskselectionFLAME3datasetUAVmonitoring
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

The paper claims that a drone can estimate per-pixel wildfire temperatures from ordinary RGB images alone, as long as a multimodal teacher network has first learned from paired RGB-thermal data and radiometric TIFF ground truth. The authors build SAM-TIFF, a teacher-student distillation framework in which an RGB-thermal fusion network supervises an RGB-only U-Net or DeepLabV3+ student that simultaneously segments fire regions and regresses temperature. On the FLAME 3 test split, the best student reaches about 68% of fire-region pixels within ±25°C and 84% within ±50°C, with a segmentation mIoU of 71.4%. If correct, the result removes the need for thermal cameras at inference time, which would lower the cost, weight, and power budget of wildfire monitoring UAVs while keeping temperature information available.

What carries the argument

The load-bearing mechanism is teacher-student modality distillation with a fire-region-masked temperature loss. The teacher is an RGB-thermal fusion network, such as SFAFMA with a ResNet-50 backbone, pretrained on FLAME 2 and fine-tuned on FLAME 3; the student is a U-Net or DeepLabV3+ that sees RGB input only. Distillation transfers the teacher's segmentation predictions to the student, while the region-masked L1 loss, $L_{\text{Flame-L1}}$, supervises temperature regression only on pixels flagged as fire by SAM pseudo-labels, forcing the student to learn an RGB-to-temperature mapping. The pseudo-labels themselves come from SAM point prompts guided by Canny edge detection and Otsu thresholding, with TOPSIS selecting among SAM's candidate masks using thermal criteria such as IoU with thresholded thermal JPGs, temperature patch means, and SSIM.

What would settle it

Have human annotators draw fire boundaries on a held-out subset of FLAME 3, then recompute the student's ±25°C and ±50°C temperature accuracy only inside the human masks and compare with the same accuracy computed inside the SAM pseudo-label masks; a large gap would show the reported numbers are artifacts of the auto-generated labels.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that thermal knowledge can be transferred into an RGB-only model well enough to recover continuous temperature fields, not just fire/no-fire labels. SAM-TIFF trains a multimodal teacher on paired RGB-Thermal frames and FLAME 3's calibrated radiometric TIFFs, then distills its segmentation predictions into a student that sees only RGB. The student's temperature head uses sigmoid scaling to constrain outputs to 0–500°C and is supervised by a region-masked L1 loss, $L_{\text{Flame-L1}}$, that computes absolute error only on pixels flagged as fire by SAM-generated pseudo-labels. The authors report that across teacher-student variants, fire-region temperature accuracy clusters around 68% within ±25°C and 83–85% within ±50°C, with the SFAFMA-50 teacher plus DeepLabV3+ student giving the best balance (71.4% mIoU, 47.3% fire IoU). They position this as the first per-pixel wildfire temperature regression from RGB UAV data, enabled by the first use of radiometric TIFFs as temperature ground truth.

Load-bearing premise

The evaluation treats SAM-generated pseudo-labels as the ground-truth fire region for both training and temperature scoring; if those masks do not match true fire boundaries, the reported accuracy and IoU numbers measure the model against its own mask generator, not against real fire regions.

Editorial extensions

If this is right

  • Deployed fire-monitoring UAVs can drop thermal sensors entirely while still outputting per-pixel temperature maps, reducing payload, cost, and power draw.
  • RGB-only temperature regression makes existing RGB drone fleets and even consumer cameras potential wildfire temperature sensors via a software update.
  • The SAM-guided pseudo-label pipeline turns unlabeled thermal datasets into segmentation supervision, removing the manual annotation bottleneck for fire masks.
  • Radiometric TIFFs become a reusable ground-truth modality for training temperature-aware models beyond classification, opening the door to hotspot and heat-gradient analysis.
  • Because the teacher can be any RGB-thermal fusion network, the distillation recipe is architecture-agnostic; better teachers or fusion blocks should raise the student ceiling.

Reading between the lines

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

  • The reported accuracy is measured only inside the SAM pseudo-label fire regions; if those masks are conservative or leaky, the same model evaluated on human-drawn fire boundaries could score noticeably lower, an evaluation the paper does not run.
  • At 68% within ±25°C, the student is far coarser than the ±2°C precision of high-end drone IR sensors, so RGB-only inference is better read as a coarse hotspot detector than a thermal-camera replacement; closing that gap is the natural next target.
  • A stress test worth running is deployment on smoke-obscured or nighttime RGB frames, where color cues degrade; since the student never sees thermal input, the distillation can only help to the extent the teacher's thermal channel reveals what RGB hides.
  • The 0–500°C clipping and the small saturated-pixel population mean predictions above 450–500°C are untrustworthy; an extension would be a separate saturation-aware head or uncertainty estimate for extreme temperatures.
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

5 major / 6 minor

Summary. The paper proposes SAM-TIFF, a teacher-student distillation framework for joint wildfire segmentation and per-pixel temperature regression from RGB-only input. A multimodal RGB-Thermal teacher network is pretrained on FLAME 2, then fine-tuned on FLAME 3 paired with automatically generated SAM/TOPSIS pseudo-masks; an RGB-only student (U-Net or DeepLabV3+) is trained to imitate the teacher's segmentation outputs and regress temperature against radiometric TIFF values via a region-masked L1 loss. On an 80-20 random split of 1,512 FLAME 3 images, the best model achieves 71.44% mIoU, 47.28% fire-class IoU, and about 68-69% of fire-region pixels within ±25°C and 84-85% within ±50°C. The paper claims to be the first to perform per-pixel wildfire temperature inference from RGB data alone and claims strong generalization on FLAME 3.

Significance. If the results were credible, this would be a valuable contribution: it targets a practical deployment problem (thermal-sensor-free wildfire monitoring) and makes use of the new radiometric TIFF ground truth in FLAME 3, an interesting data source that the community is only beginning to exploit. The teacher-student design, the SAM-based pseudo-labeling pipeline, and the comparison across multiple teacher architectures are useful elements. However, the significance hinges entirely on whether the reported temperature and segmentation numbers are meaningful. As detailed in the major comments, the current evaluation is compromised by shared pseudo-labels between training and testing, missing baseline comparisons, and unsupported generalization claims. These issues are load-bearing, so the contribution, while promising, is not yet established.

major comments (5)
  1. [Section V and Section VII, Eq. (2), Tables III and IV] The evaluation is circular with respect to the pseudo-labels. Section V constructs fire-region masks using SAM prompted by thermal-derived Canny/Otsu points and selects masks via TOPSIS criteria that include IoU against a thresholded thermal JPG, Otsu IoU, and temperature patch means. Section VII then uses these same pseudo-labels as the evaluation reference: Table III computes IoU against the SAM masks, and Table IV restricts temperature evaluation to pixels where the SAM mask equals 1. In addition, Eq. (2) masks the temperature training loss to F={i|SAM_i=1}. Consequently, the student is never required to detect fire boundaries independently before temperature accuracy is scored, and any fire pixels missed by the auto-mask are excluded from both segmentation and temperature metrics. A model that fits SAM-positive pixels well can report ~68% within ±25°C and ~84% within ±50°C while performing arbitrarily poorly on the true fire region. The manual review in Section V only pruned low-quality masks; it does not provide independent annotations for scoring. This circularity affects the central numerical claims of the paper, not a side experiment. The authors should evaluate on independent ground-truth fire boundaries (e.g., manually annotated masks, or at least full-image temperature metrics) and report results over all pixels or over a held-out annotation set.
  2. [Section VII, Tables III and IV] The paper claims that distillation from a multimodal teacher enables RGB-only temperature inference, but it never compares against the natural baseline: an RGB-only student trained directly on the RGB input with the same masked L1 temperature loss (and, ideally, the same segmentation objective) without teacher supervision. Without such a non-distilled baseline, it is impossible to tell whether the reported temperature accuracy is due to the distillation, to the pseudo-label masks, or simply to the RGB-to-temperature regression being learnable directly. The absence of this ablation is a significant omission for a paper whose central claim is about the benefit of multimodal distillation.
  3. [Abstract and Section VII] The claim of 'strong generalization on the recent FLAME 3 dataset' is unsupported. The 80-20 random split described in Section VII likely mixes burn locations in both training and testing, so the test set may contain frames from the same burns as the training set. No cross-location experiment is reported (e.g., training on Shoetank and Sycan2A/2D and testing on Willamette Valley, or leave-one-burn-out cross-validation). The paper's own discussion in Section VII notes that some burns (e.g., Willamette Valley) are more consistent and provide better results, which suggests a strong location effect. A meaningful generalization claim requires either a cross-location evaluation or an explicit discussion of why a random split is appropriate. As is, the results may reflect overfitting to burn-specific conditions rather than generalizable RGB-only perception.
  4. [Section VII (experimental setup) and Tables III and IV] All results are reported without error bars or multiple training runs, which is problematic given the small dataset (1,512 images, 80-20 split yields roughly 300 test images). The absence of repeated seeds makes it impossible to assess the variance of the reported metrics. Additionally, the comparison across teachers is inconsistent: Section VII states that EAEF-50 converged faster and is reported at epoch 150, while all other models are reported at epoch 270. This introduces a confound—differences in Table III and Table IV across teacher rows may be due to training duration rather than teacher quality. The authors should report mean and standard deviation over at least several seeds and use a consistent training protocol (or justify the early stopping for EAEF-50 with a validation metric).
  5. [Equations (1)-(3) and Section V] Several free hyperparameters are not reported, making the method non-reproducible. Specifically, λDice is set to 0.5, but the values of λstudent_dice and λFlame-L1 in Eq. (3) are not given; the temperature clipping range (stated as 0–500°C in Section II.C) is not tied to where it is applied in the pipeline; the Canny high threshold appears only in the Figure 2 caption as 200°C; the error margin ε is given as 25°C in the caption but not in the text; and the TOPSIS criterion weights are described qualitatively (thermal IoU 'given greater weight') but never specified quantitatively. A hyperparameter table and a precise description of the mask-generation thresholds are necessary for others to reproduce or build on this work.
minor comments (6)
  1. [Section II.C] There are typos: 'an approach top unify' should be 'an approach to unify', and 'low quanity' should be 'low quantity'.
  2. [Figure 2 caption and Section V.A] The notation for Canny thresholds is confusing. The caption says 'Low Threshold = τ and High Threshold = 200 degrees Celsius', while the text defines τ as the Otsu threshold computed per image. Clarify whether the Canny low threshold is indeed the Otsu threshold or a separate parameter, and whether τ is the same variable in both places.
  3. [Section VI] The description of the FLAME 2 split is ambiguous: 'The full dataset train-validation-test split was 80-10-10 and then 10000 training, 1000 validation, and 1000 testing images were randomly sampled from this.' Clarify whether the 80-10-10 split is performed first and then samples are drawn, or whether the samples are drawn from the full dataset and then split. Also, the same seed being used 'for each network training' does not guarantee the same split if the sampling occurs before network training; this should be described precisely.
  4. [Tables II and III] Percentage values are reported inconsistently: Table II lists mIoU as 0.9563 but the text reports '96.69% mIoU', and Table III lists IoU values as 0.9515 etc. while the text reports them as percentages. Choose one format (decimal or percentage) and use it consistently throughout the tables and text.
  5. [Section VII (temperature evaluation)] The temperature accuracy metric is reported only for fire-region pixels (pixels where the SAM mask equals 1). This ignores both false positives (predicted fires on non-fire pixels) and false negatives (missed fires), which are exactly the pixels where an RGB-only model could fail dangerously in practice. Reporting full-image temperature errors or per-class breakdowns would make the practical implications clearer.
  6. [Section VI, paragraph on FLAME 2 pretraining] The paper notes that the high FLAME 2 numbers likely reflect the networks learning the Otsu thresholding and thermal colormap rather than true fire segmentation. This is an important caveat, but it is only mentioned in one sentence and is not carried into the discussion of the final FLAME 3 results. Please discuss how this pretraining bias might propagate through the distillation and affect the final temperature predictions.

Circularity Check

3 steps flagged · score 4.0 of 10

Segmentation and fire-region temperature metrics are scored against the same SAM/TOPSIS pseudo-labels used to define the training region, so the reported mIoU and ±25/±50°C figures are partly self-referential; the underlying TIFF temperature regression is not itself circular.

  1. self definitional [Section V ('Final Ground Truth Mask Decision'); Section VII (Table III)]
    "This approach enables objective mask selection by evaluating each candidate based on five quantitative criteria... IoU (Thresholded Thermal JPG vs. SAM): Quantifies the overlap between the SAM mask and a thresholded thermal image. Higher values indicate better alignment with thermal features and are thus prioritized... The final mask selected is the one with the highest TOPSIS score across all criteria."

    The masks are called the 'Final Ground Truth Mask' but are produced by a pipeline whose TOPSIS selection criteria are thermal-derived: IoU with a thresholded thermal JPG, IoU with Otsu masks, and mean foreground temperature differences from radiometric TIFF data. Section VII then scores the student's segmentation IoU against these same auto-generated masks. The reported 71.44% mIoU and 47.28% fire-class IoU therefore measure how well the student reproduces the pipeline's own thermal-guided mask choices, not agreement with an independent fire-region annotation. The manual review step pruned low-quality masks but did not create a separate scoring reference, so the segmentation evaluation is self-referential.

  2. fitted input called prediction [Section IV, Eq. (2); Section VII, Table IV]
    "F denotes the set of pixel indices corresponding to fire regions, defined as F ={i|SAMi = 1}... For the fire region pixels, the SAM pseudo label fire region pixel locations were extracted and then those locations were used to compare the predicted student network temperatures to the ground truth TIFF temperatures."

    Equation (2) restricts the temperature training loss to F = {i | SAM_i = 1}, and Table IV extracts exactly 'the SAM pseudo label fire region pixel locations' for evaluation. The same auto-generated mask therefore defines both the supervision target and the evaluation region. Any true fire pixels missed by the SAM/TOPSIS pipeline are excluded from both the loss and the metric, so a student can report ~68% within ±25°C and ~84% within ±50°C by fitting SAM-positive pixels while performing arbitrarily on fire pixels outside the mask. The radiometric TIFF values themselves are independent, so this is not a fitted-constant temperature prediction, but the reported fire-region accuracy is conditional on the method's own label-generation step.

1 more flagged steps
  1. other [Section VI, 'FLAME 2 Pretraining' (discussion after Table II)]
    "It is important to note that these results indicate some degree of overfitting, most likely caused by the ground truth being generated from Otsu's Method directly from the thermal JPGs themselves. This lends itself to these networks potentially learning the thresholding of Otsu's Method or thermal colormapping during training rather than the true segmentation of the fire."

    The paper itself concedes that the FLAME 2 pretraining labels were Otsu masks computed directly from thermal JPGs and that the high scores may reflect learning the thresholding algorithm rather than true fire structure. Those teacher checkpoints are loaded as the teachers for SAM-TIFF distillation, so any Otsu-thresholding bias is inherited by the teacher and passed to the student. This is a self-admitted circularity in the teacher's pretraining supervision, though it is secondary to the FLAME 3 results because the final temperature regression is still evaluated against independent radiometric TIFF values on the chosen mask region.

full rationale

The core temperature regression is not circular: the regression targets T_i are independent radiometric TIFF values in degrees Celsius, and the student temperature head is trained with an L1 loss against those TIFF values, not against a fitted constant or a renaming of the RGB input. However, the fire region used in both training and evaluation is the same SAM/TOPSIS pseudo-label mask. That mask is generated with thermal-derived criteria, including IoU with a thresholded thermal JPG, Otsu masks, and temperature patch means, so the evaluation reference for both segmentation and fire-region temperature accuracy is not independent of the thermal modality the student is meant to emulate. Section VII explicitly extracts the SAM pseudo-label fire-region pixel locations for the temperature comparison, which is the same set F used in Eq. (2) to mask the temperature loss. This means any fire pixels omitted by the pseudo-label pipeline are invisible in both training and evaluation, so the reported mIoU and ±25/±50°C percentages are partly self-referential. The paper also discloses that the FLAME 2 pretraining masks came from Otsu thresholding of thermal JPGs and that the networks may have learned that thresholding rather than true fire segmentation, which further weakens the teacher chain. No load-bearing self-citation circularity was found: FLAME 3 is a public dataset and the temperature ground truth is external. The manual review of masks is a quality filter, not an independent annotation set. Overall, the central numerical claims are affected by a self-defined evaluation region, but the temperature regression itself retains independent content, so a score of 4 is appropriate.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the reliability of pseudo-labels derived from the same thermal modality the student tries to replace. The free parameters are mostly training hyperparameters and mask-generation thresholds, several of which are not reported. No new physical entities are introduced.

free parameters (7)
  • λDice = 0.5
    Weight for dice loss in teacher loss (Eq. 1); chosen by hand, no sensitivity analysis.
  • λstudent_dice = not reported
    Weight for student dice loss (Eq. 3); value not given in the paper.
  • λFlame-L1 = not reported
    Weight for temperature L1 loss (Eq. 3); value not given in the paper.
  • Temperature clipping range = 0 to 500 °C
    TIFF values clipped to this range to remove sky and saturated pixels; affects the regression target.
  • Canny high threshold = 200 °C
    Used in autopoint locator to place point prompts near fire edges.
  • Error margin ϵ = 25 °C
    Applied to Otsu threshold to counteract foreground misclassification.
  • TOPSIS criteria weights = not fully specified
    Thermal IoU given 'greater weight', but exact weights not listed.
assumptions (4)
  • domain assumption FLAME 3 radiometric TIFFs provide accurate per-pixel ground truth temperatures.
    Taken from dataset paper [4]; no independent calibration verification in this work.
  • domain assumption SAM-generated masks selected by TOPSIS are valid fire segmentation ground truth.
    Used as supervision and evaluation masks; no human-annotated validation set.
  • domain assumption Otsu thresholding of FLAME 2 thermal JPGs produces adequate segmentation targets for teacher pretraining.
    The authors admit this likely taught networks to imitate Otsu thresholding/colormapping rather than true fire shape.
  • domain assumption Thermal modality is informative for fire segmentation and can be distilled into an RGB-only student.
    The premise of cross-modal distillation; not independently tested with ablations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Seeing Heat with Color -- RGB-Only Wildfire Temperature Inference from SAM-Guided Multimodal Distillation using Radiometric Ground Truth." pith.science (2026). https://pith.science/paper/WDPLL3KS

@misc{pith2026250501638,
  author       = {Pith},
  title        = {Pith review of: Seeing Heat with Color -- RGB-Only Wildfire Temperature Inference from SAM-Guided Multimodal Distillation using Radiometric Ground Truth},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WDPLL3KS}},
  note         = {Machine review of arXiv:2505.01638}
}
read the original abstract

High-fidelity wildfire monitoring using Unmanned Aerial Vehicles (UAVs) typically requires multimodal sensing - especially RGB and thermal imagery - which increases hardware cost and power consumption. This paper introduces SAM-TIFF, a novel teacher-student distillation framework for pixel-level wildfire temperature prediction and segmentation using RGB input only. A multimodal teacher network trained on paired RGB-Thermal imagery and radiometric TIFF ground truth distills knowledge to a unimodal RGB student network, enabling thermal-sensor-free inference. Segmentation supervision is generated using a hybrid approach of segment anything (SAM)-guided mask generation, and selection via TOPSIS, along with Canny edge detection and Otsu's thresholding pipeline for automatic point prompt selection. Our method is the first to perform per-pixel temperature regression from RGB UAV data, demonstrating strong generalization on the recent FLAME 3 dataset. This work lays the foundation for lightweight, cost-effective UAV-based wildfire monitoring systems without thermal sensors.

Figures

Figures reproduced from arXiv: 2505.01638 by the authors.

Figure 1
Figure 1. TIFF Temperature Range Visualization - Sycan Marsh [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. SAM-TIFF Architecture (a), Autpoint Locator Algorithm (b) - ‘TIFF Guided Point Prompts’ Module in (a); Jet [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Willamette Valley Sample Image Results for Teacher [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figures from the paper (1 more)
Figure 3
Figure 3. Figure 3: Segmentation predictions for SFAFMA-50 Teacher - [PITH_FULL_IMAGE:figures/full_fig_p006_3.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 13 canonical work pages

  1. [1]

    A comprehensive survey of research towards ai-enabled unmanned aerial systems in pre-, active-, and post-wildfire manage- ment,

    S. P. H. Boroujeni, A. Razi, S. Khoshdel, F. Afghah, J. L. Coen, L. O’Neill, P. Fule, A. Watts, N.-M. T. Kokolakis, and K. G. Vamvoudakis, “A comprehensive survey of research towards ai-enabled unmanned aerial systems in pre-, active-, and post-wildfire manage- ment,” Information Fusion , vol. 108, p. 102369, 2024

    work page 2024

  2. [2]

    Thermal image calibration and correction using unpaired cycle-consistent adversarial networks,

    H. Rajoli, P. Afshin, and F. Afghah, “Thermal image calibration and correction using unpaired cycle-consistent adversarial networks,” in2023 57th Asilomar Conference on Signals, Systems, and Computers , 2023, pp. 1425–1429

    2023

  3. [3]

    FlameFinder: Illuminating Obscured Fire through Smoke with Attentive Deep Metric Learning

    H. Rajoli, S. Khoshdel, F. Afghah, and X. Ma, “Flamefinder: Illuminating obscured fire through smoke with attentive deep metric learning,” 2024. [Online]. Available: https://arxiv.org/abs/2404.06653

    work page Pith review arXiv 2024

  4. [4]

    Flame 3 - radiometric thermal uav imagery for wildfire management,

    B. Hopkins, L. O’Neill, M. Marinaccio, F. Afghah, E. Rowell, R. Parsons, and S. Flanary, “Flame 3 - radiometric thermal uav imagery for wildfire management,” 2024. [Online]. Available: https://dx.doi.org/10.21227/w0mz-aq48

    work page doi:10.21227/w0mz-aq48 2024

  5. [5]

    Flame 3 dataset: Unleashing the power of radiometric thermal uav imagery for wildfire management,

    B. Hopkins, L. ONeill, M. Marinaccio, E. Rowell, R. Parsons, S. Flanary, I. Nazim, C. Seielstad, and F. Afghah, “Flame 3 dataset: Unleashing the power of radiometric thermal uav imagery for wildfire management,”

    work page

  6. [6]

    Computer vision for wildfire research: An evolving image dataset for processing and analysis,

    T. Toulouse, L. Rossi, A. Campana et al., “Computer vision for wildfire research: An evolving image dataset for processing and analysis,” Fire Safety Journal, vol. 92, pp. 188–194, 2017

    work page 2017

  7. [7]

    Deep-learning- based aerial image classification for emergency response applications using unmanned aerial vehicles,

    Kyrkou, Christos and Theocharides, Theocharis, “Deep-learning- based aerial image classification for emergency response applications using unmanned aerial vehicles,” 2019. [Online]. Available: https: //arxiv.org/abs/1906.08716

    arXiv 2019

  8. [8]

    Wildfire detection image data,

    Dincer, Baris, “Wildfire detection image data,” 2021. [Online]. Available: https://www.kaggle.com/datasets/brsdincer/ wildfire-detection-image-data?resource=download

    2021

Show all 27 references
  1. [9]

    Fire and smoke dataset,

    DataCluster Labs, “Fire and smoke dataset,” 2021. [Online]. Available: https://www.kaggle.com/datasets/dataclusterlabs/fire-and-smoke-dataset

    2021

  2. [10]

    Fire and smoke datasets in 20 years: An in-depth review,

    S. P. H. Boroujeni, N. Mehrabi, F. Afghah, C. P. McGrath, D. Bhatkar, M. A. Biradar, and A. Razi, “Fire and smoke datasets in 20 years: An in-depth review,” 2025. [Online]. Available: https://arxiv.org/abs/2503.14552

    2025 arXiv

  3. [11]

    The flame dataset: Aerial imagery pile burn detection using drones (uavs),

    A. Shamsoshoara, F. Afghah, A. Razi, L. Zheng, P. Ful ´e, and E. Blasch, “The flame dataset: Aerial imagery pile burn detection using drones (uavs),” 2020. [Online]. Available: https://dx.doi.org/10.21227/ qad6-r683

    2020

  4. [12]

    Flame 2: Fire detection and modeling: Aerial multi-spectral image dataset,

    B. Hopkins, L. O’Neill, F. Afghah, A. Razi, E. Rowell, A. Watts, P. Fule, and J. Coen, “Flame 2: Fire detection and modeling: Aerial multi-spectral image dataset,” 2022. [Online]. Available: https://dx.doi.org/10.21227/swyw-6j78

    2022 doi

  5. [13]

    Encoder-decoder with atrous separable convolution for semantic image segmentation,

    L.-C. Chen, Y . Zhu, G. Papandreou, F. Schroff, and H. Adam, “Encoder-decoder with atrous separable convolution for semantic image segmentation,” 2018. [Online]. Available: https://arxiv.org/abs/1802. 02611

    2018

  6. [14]

    Mfnet: Towards real-time semantic segmentation for autonomous vehicles with multi-spectral scenes,

    Q. Ha, K. Watanabe, T. Karasawa, Y . Ushiku, and T. Harada, “Mfnet: Towards real-time semantic segmentation for autonomous vehicles with multi-spectral scenes,” in 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2017, pp. 5108–5115

    2017

  7. [15]

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

    Y . Sun, W. Zuo, and M. Liu, “Rtfnet: Rgb-thermal fusion network for semantic segmentation of urban scenes,” IEEE Robotics and Automation Letters, vol. 4, no. 3, pp. 2576–2583, 2019

    2019

  8. [16]

    Feanet: Feature-enhanced attention network for rgb-thermal real-time semantic segmentation,

    F. Deng, H. Feng, M. Liang, H. Wang, Y . Yang, Y . Gao, J. Chen, J. Hu, X. Guo, and T. L. Lam, “Feanet: Feature-enhanced attention network for rgb-thermal real-time semantic segmentation,” 2021. [Online]. Available: https://arxiv.org/abs/2110.08988

    2021 arXiv

  9. [17]

    Sfaf-ma: Spatial feature aggregation and fusion with modality adaptation for rgb-thermal semantic segmentation,

    X. He, M. Wang, T. Liu, L. Zhao, and Y . Yue, “Sfaf-ma: Spatial feature aggregation and fusion with modality adaptation for rgb-thermal semantic segmentation,” IEEE Transactions on Instrumentation and Measurement, vol. 72, pp. 1–10, 2023

    2023

  10. [18]

    A real-time fire segmentation method based on a deep learning approach,

    M. Li, Y . Zhang, L. Mu, J. Xin, Z. Yu, S. Jiao, H. Liu, G. Xie, and Y . Yingmin, “A real-time fire segmentation method based on a deep learning approach,” IF AC-PapersOnLine, vol. 55, no. 6, pp. 145–150, 2022, 11th IFAC Symposium on Fault Detection, Supervision and Safety for...

    2022

  11. [19]

    Detecting wildfires on uavs with real-time segmentation trained by larger teacher models,

    J. Pesonen, T. Hakala, V . Karjalainen, N. Koivum ¨aki, L. Markelin, A.-M. Raita-Hakola, J. Suomalainen, I. P ¨ol¨onen, and E. Honkavaara, “Detecting wildfires on uavs with real-time segmentation trained by larger teacher models,” 2024. [Online]. Available: https://arxiv.org/a...

    2024 arXiv

  12. [20]

    Segment anything,

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Lo, P. Doll ´ar, and R. Girshick, “Segment anything,” 2023. [Online]. Available: https: //arxiv.org/abs/2304.02643

    2023 arXiv

  13. [21]

    Hwang and K

    C.-L. Hwang and K. Yoon, Methods for Multiple Attribute Decision Making. Berlin, Heidelberg: Springer Berlin Heidelberg, 1981, pp. 58–

    1981

  14. [22]

    A threshold selection method from gray-level histograms,

    N. Otsu, “A threshold selection method from gray-level histograms,” IEEE Transactions on Systems, Man, and Cybernetics , vol. 9, no. 1, pp. 62–66, 1979

    1979

  15. [23]

    Wildland fire detection and monitoring using a drone-collected rgb/ir image dataset,

    X. Chen, B. Hopkins, H. Wang, L. O’Neill, F. Afghah, A. Razi, P. Ful´e, J. Coen, E. Rowell, and A. Watts, “Wildland fire detection and monitoring using a drone-collected rgb/ir image dataset,” IEEE Access , vol. 10, pp. 121 301–121 317, 2022

    2022

  16. [24]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” 2015. [Online]. Available: https://arxiv.org/abs/1505.04597

    2015 arXiv

  17. [25]

    Explicit attention-enhanced fusion for rgb-thermal perception tasks,

    M. Liang, J. Hu, C. Bao, H. Feng, F. Deng, and T. L. Lam, “Explicit attention-enhanced fusion for rgb-thermal perception tasks,” 2023. [Online]. Available: https://arxiv.org/abs/2303.15710

    2023 arXiv

  18. [191]

    Available: https://doi.org/10.1007/978-3-642-48318-9 3

    [Online]. Available: https://doi.org/10.1007/978-3-642-48318-9 3

    doi

  19. [2024]

    Available: https://arxiv.org/abs/2412.02831

    [Online]. Available: https://arxiv.org/abs/2412.02831

    arXiv

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.