A Comparative Performance Evaluation of YOLO-Type Detectors on a New Open Fire and Smoke Dataset
Abstract
:1. Introduction
- The generation of a novel large open dataset that includes challenging images that may (but should not) be detected as fire instances;
- A two-step annotation procedure that combines an initial automatic solution based on the use of the YOLO8 detection model with a second manual correction and bounding box adjustment phase that also removes labeling errors;
- A performance comparison of seven YOLO models, including the most recent YOLO10 member of this family of object detectors;
- A comparative analysis of the detection efficiency of the analyzed models for daylight vs. night-time images.
2. Datasets and Models for Fire/Smoke Identification
2.1. Fire and Smoke Datasets
2.2. Machine Learning Models for Fire/Smoke Identification
3. Protocol for Dataset Generation
3.1. Dataset Composition
- (a)
- Daylight scenes (15,812 images):
- -
- 2121 smoke-only scenes (thin or dense, small/medium/large size, far/medium/close distance)
- -
- 2475 fire-only scenes (small/medium/large size, far/medium/close distance, car/building/trash/wood burning)
- -
- 645 other-only scenes (different types of clouds, sunrises/sunsets, car headlights, or other types of shiny surfaces)
- -
- 4312 fire-and-smoke scenes (small/medium/large size, far/medium/close distance, car/building/trash/wood burning)
- -
- 3138 drone-captured scenes containing fire-and-smoke or just smoke-only images (small/medium/large size, far/medium/close distance, car/building/trash/wood burning)
- -
- 3121 vegetation fires (forest fires, grass fires, wildfires)
- (b)
- Night-time scenes (7158 images):
- -
- 2657 fire-and-smoke scenes (small/medium/large size, far/medium/close distance, car/building/wood burning)
- -
- 1852 fire, smoke, and other scenes (small/medium/large size, far/medium/close distance, car/building burning, car lights, streetlamps, illuminated advertisement panels)
- -
- 1228 fire-only scenes (small/medium/large size, far/medium/close distance, car/building burning)
- -
- 1421 other-only scenes (small/medium/large size, far/medium/close distance, car lights, streetlamps, illuminated advertisement panels)
3.2. Data Labeling Procedure
4. Fire/Smoke Detection Evaluation
5. Discussion
System | Model | Precision | Recall | F1 | mAP@50 | Dataset Size |
---|---|---|---|---|---|---|
Tian [52] | YOLOv5n | 0.92 | 0.91 | 0.91 | 0.94 | >10,000 |
Avazov [53] | YOLOv7 | 0.97 | 0.94 | 0.96 | 0.97 | 4622 |
Talaat [45] | YOLOv8 | 0.93 | 0.94 | 0.94 | 0.97 | >25,000 |
Luo [31] | YOLOX | 0.86 | 0.72 | 0.78 | 0.84 | 5000 |
Geng [54] | YOLOFM | 0.94 | 0.94 | 0.94 | 0.97 | 18,644 |
This work | YOLO10 | 0.86 | 0.85 | 0.85 | 0.91 | >22,000 |
6. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- Kumar, A. Preserving life on earth. In Adaptation, Ecosystem-Based; Elsevier: Amsterdam, The Netherlands, 2022; pp. 503–602. [Google Scholar]
- International Association of Fire and Rescue Services. Available online: www.ctif.org/news/world-fire-statistics-report-no-29-comprehensive-analysis-fire-statistics-2022 (accessed on 2 July 2024).
- Zhang, F.; Zhao, P.; Xu, S.; Wu, Y.; Yang, X.; Zhang, Y. Integrating multiple factors to optimize watchtower deployment for wildfire detection. Sci. Total Environ. 2020, 737, 139561. [Google Scholar] [CrossRef]
- Chagger, R.; Smith, D. The Causes of False Fire Alarms in Buildings; Tech. Rep. BC2982, no. 1; BRE Trust: Hertfordshire, UK, 2014. [Google Scholar]
- Statistics 2008; Duisburg Fire Brigade: Duisburg, Germany, 2009.
- Kruell, W.; Schultze, T.; Tobera, R.; Willms, I. Analysis of dust properties to solve the complex problem of non-fire sensitivity testing of optical smoke detectors. Procedia. Eng. 2013, 62, 859–867. [Google Scholar] [CrossRef]
- Saleh, A.; Zulkifley, M.A.; Harun, H.H.; Gaudreault, F.; Davison, I.; Spraggon, M. Forest fire surveillance systems: A review of deep learning methods. Heliyon 2024, 10, E23127. [Google Scholar] [CrossRef] [PubMed]
- Wilkinson, M.D.; Dumontier, M.; Jan Aalbersberg, I.J.; Appleton, G.; Axton, M.; Baak, A.; Blomberg, N.; Boiten, J.-W.; da Silva Santos, L.B.; Bourne, P.E.; et al. The FAIR guiding principles for scientific data management and stewardship. Sci. Data 2016, 3, 160018. [Google Scholar] [CrossRef]
- Fire and Smoke Dataset. Available online: https://thub.com/CostiCatargiu/NEWFireSmokeDataset_YoloModels (accessed on 26 July 2024).
- Terven, J.; Córdova-Esparza, D.M.; Romero-González, J.A. A comprehensive review of YOLO architectures in computer vision: From YOLOv1 to YOLOv8 and YOLO-NAS. Mach. Learn. Knowl. Extr. 2023, 5, 1680–1716. [Google Scholar] [CrossRef]
- Foggia, P.; Saggese, A.; Vento, M. Real-Time fire detection for video-surveillance applications using a combination of experts based on color, shape, and motion. IEEE Trans. Circ. Syst. Video Technol. 2015, 25, 1545–1556. [Google Scholar] [CrossRef]
- Li, S.; Yan, Q.; Liu, P. An Efficient fire detection method based on multiscale feature extraction, implicit deep supervision and channel attention mechanism. IEEE Trans. Image Process. 2020, 29, 8467–8475. [Google Scholar] [CrossRef] [PubMed]
- Yar, H.; Hussain, T.; Agarwal, M.; Khan, Z.A.; Gupta, S.K.; Wook, S. Baik Optimized dual fire attention network and medium-scale fire classification benchmark. IEEE Trans. Image Process. 2022, 31, 6331–6343. [Google Scholar] [CrossRef]
- Dilshad, N.; Khan, S.U.; Alghamdi, N.S.; Taleb, T.; Song, J. Toward efficient fire detection in IoT environment: A modified attention network and large-scale data set. IEEE Internet Things J. 2024, 11, 13467–13481. [Google Scholar] [CrossRef]
- Shamsoshoara, A.; Afghah, F.; Razi, A.; Zheng, L.; Fulé, P.Z.; Blasch, E. Aerial imagery pile burn detection using deep learning: The FLAME dataset. Comput. Netw. 2021, 193, 108001. [Google Scholar] [CrossRef]
- Yar, H.; Ullah, W.; Khan, Z.A.; Baik, S.W. An effective attention-based CNN model for fire detection in adverse weather conditions. ISPRS J. Photogramm. Remote Sens. 2023, 206, 335–346. [Google Scholar] [CrossRef]
- Wang, M.; Yue, P.; Jiang, L.; Yu, D.; Tuo, T.; Li, J. An open flame and smoke detection dataset for deep learning in remote sensing based fire detection. Geo-Spat. Inf. Sci. 2024, 1–16. [Google Scholar] [CrossRef]
- Çelik, T.; Demirel, H. Fire detection in video sequences using a generic color model. Fire Saf. J. 2009, 44, 147–158. [Google Scholar] [CrossRef]
- Poobalan, K.; Liew, S.-C. Fire detection based on color filters and Bag-of-Features classification. In Proceedings of the 2015 IEEE Student Conference on Research and Development, Kuala Lumpur, Malaysia, 13–14 December 2015. [Google Scholar]
- Yu, C.; Fang, J.; Wang, J.; Zhang, Y. Video fire smoke detection using motion and color features. Fire Technol. 2010, 46, 651–663. [Google Scholar]
- Cui, Y.; Dong, H.; Zhou, E. An early fire detection method based on smoke texture analysis and discrimination. In Proceedings of the 2008 Congress on Image and Signal Processing, Sanya, China, 27–30 May 2008. [Google Scholar]
- Yu, C.; Zhang, Y.; Fang, J.; Wang, J. Texture analysis of smoke for real-time fire detection. In Proceedings of the 2009 Second Internat Workshop on Computer Science and Engineering, Qingdao, China, 28–30 October 2009. [Google Scholar]
- Chino, D.Y.T.; Avalhais, L.P.S.; Rodrigues, J.F.; Traina, A.J.M. Bowfire: Detection of fire in still images by integrating pixel color and texture analysis. In Proceedings of the 2015 28th SIBGRAPI Conference on Graphics, Patterns and Images, Salvador, Brazil, 26–29 August 2015. [Google Scholar]
- Ye, W.; Zhao, J.; Wang, S.; Wang, Y.; Zhang, D.; Yuan, Z. Dynamic texture based smoke detection using Surfacelet transform and HMT model. Fire Saf. J. 2015, 73, 91–101. [Google Scholar] [CrossRef]
- Toulouse, T.; Rossi, L.; Celik, T.; Akhloufi, M. Automatic fire pixel detection using image processing: A comparative analysis of rule-based and machine learning-based methods. Signal Image Video Process. 2016, 10, 647–654. [Google Scholar] [CrossRef]
- Avgeris, M.; Spatharakis, D.; Dechouniotis, D.; Kalatzis, N.; Roussaki, I.; Papavassiliou, S. Where there is fire there is SMOKE: A scalable edge computing framework for early fire detection. Sensors 2019, 19, 639. [Google Scholar] [CrossRef] [PubMed]
- Zhang, Z.; Zhao, J.; Zhang, D.; Qu, C.; Ke, Y.; Cai, B. Contour based forest fire detection using FFT and wavelet. In Proceedings of the 2008 International Conference on Computer Science and Software Engineering, Wuhan, China, 12–14 December 2008. [Google Scholar]
- Dimitropoulos, K.; Barmpoutis, P.; Grammalidis, N. Spatio-temporal flame modeling and dynamic texture analysis for automatic video-based fire detection. IEEE Trans. Circuits Syst. Video Technol. 2015, 25, 339–351. [Google Scholar] [CrossRef]
- Prema, C.E.; Vinsley, S.S.; Suresh, S. Efficient flame detection based on static and dynamic texture analysis in forest fire detection. Fire Technol. 2018, 54, 255–288. [Google Scholar] [CrossRef]
- Ba, R.; Chen, C.; Yuan, J.; Song, W.; Lo, S. SmokeNet: Satellite smoke scene detection using convolutional neural network with spatial and channel-wise attention. Remote Sens. 2019, 11, 1702. [Google Scholar] [CrossRef]
- Luo, M.; Xu, L.; Yang, Y.; Cao, M.; Yang, J. Laboratory flame smoke detection based on an improved YOLOX algorithm. Appl. Sci. 2022, 12, 12876. [Google Scholar] [CrossRef]
- Zhang, Q.X.; Lin, G.H.; Zhang, Y.M.; Xu, G.; Wang, J.J. Wildland forest fire smoke detection based on Faster R-CNN using synthetic smoke images. Procedia Eng. 2018, 211, 441–446. [Google Scholar] [CrossRef]
- Chaoxia, C.; Shang, W.; Shang, W.; Zhang, F. Information-guided flame detection based on Faster R-CNN. IEEE Access 2020, 8, 58923–58932. [Google Scholar] [CrossRef]
- Tao, H.; Liu, J.; Yang, Z.; Wang, G.; Shang, J.; Qiu, H.; Gao, L. Revolutionizing flame detection: Novelization in flame detection through transferring distillation for knowledge to pruned model. Expert Syst. Appl. 2024, 249, 123787. [Google Scholar] [CrossRef]
- Tao, H.; Wang, G.; Liu, J.; Yang, Z. A deep learning-based dynamic deformable adaptive framework for locating the root region of the dynamic flames. PLoS ONE 2024, 19, e0301839. [Google Scholar] [CrossRef]
- Ghali, R.; Akhloufi, M.A.; Mseddi, W.S. Deep learning and transformer approaches for UAV-based wildfire detection and segmentation. Sensors 2022, 22, 1977. [Google Scholar] [CrossRef] [PubMed]
- Wang, C.-Y.; Yeh, I.-H.; Liao, H.-Y.M. YOLOv9: Learning What You Want to Learn Using Programmable Gradient Information. arXiv 2024, arXiv:2402.13616. [Google Scholar] [CrossRef]
- Wang, A.; Chen, H.; Liu, L.; Chen, K.; Lin, Z.; Han, J.; Ding, G. YOLOv10: Real-Time End-to-End Object Detection. arXiv 2024, arXiv:2405.14458. [Google Scholar] [CrossRef]
- YOLO-NAS Documentation. Available online: https://github.com/Deci-AI/super-gradients (accessed on 2 July 2024).
- Park, M.; Ko, B.C. Two-step real-time night-time fire detection in an urban environment using static ELASTIC-YOLOv3 and temporal fire-tube. Sensors 2020, 20, 2202. [Google Scholar] [CrossRef]
- Mukhiddinov, M.; Abdusalomov, A.B.; Cho, J. A wildfire smoke detection system using unmanned aerial vehicle images based on the optimized YOLOv5. Sensors 2022, 22, 9384. [Google Scholar] [CrossRef]
- Dou, Z.; Zhou, H.; Liu, Z.; Hu, Y.; Wang, P.; Zhang, J.; Wang, Q.; Chen, L.; Diao, X.; Li, J. An improved YOLOv5s fire detection model. Fire Technol. 2023, 60, 135–166. [Google Scholar] [CrossRef]
- Du, H.; Zhu, W.; Peng, K.; Li, W. Improved high speed flame detection method based on YOLOv7. Open J. Appl. Sci. 2022, 12, 2004–2018. [Google Scholar] [CrossRef]
- Chen, G.; Cheng, R.; Lin, X.; Jiao, W.; Bai, D.; Lin, H. LMDFS: A lightweight model for detecting forest fire smoke in UAV images based on YOLOv7. Remote Sens. 2023, 15, 3790. [Google Scholar] [CrossRef]
- Talaat, F.M.; Eldin, H.Z. An improved fire detection approach based on YOLO-v8 for smart cities. Neural. Comp. Appl. 2023, 35, 20939–20954. [Google Scholar] [CrossRef]
- Zhang, Z.; Tan, L.; Tiong, R.L.K. Ship-Fire Net: An improved YOLOv8 algorithm for ship fire detection. Sensors 2024, 24, 727. [Google Scholar] [CrossRef] [PubMed]
- Abdusalomov, A.B.; Islam, B.M.S.; Nasimov, R.; Mukhiddinov, M.; Whangbo, T.K. An Improved Forest Fire Detection Method Based on the Detectron2 Model and a Deep Learning Approach. Sensors 2023, 23, 1512. [Google Scholar] [CrossRef]
- Fire and Smoke Dataset Health Check. Available online: https://universe.roboflow.com/catargiuconstantin2/firesmokenewdataset/health (accessed on 2 July 2024).
- Petsiuk, V.; Jain, R.; Manjunatha, V.; Morariu, V.I.; Mehra, A.; Ordonez, V.; Saenko, K. Black-box explanation of object detectors via saliency maps. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Nashville, TN, USA, 20–25 June 2021; pp. 11443–11452. [Google Scholar]
- Nguyen, K.; Nguyen, T.T.H.; Nguyen, V.T.K.; Truong, V.B.; Phan, T.; Cao, H. Efficient and Concise Explanations for Object Detection with Gaussian-Class Activation Mapping Explainer. In Proceedings of the 37th Canadian Conference on Artificial Intelligence, Guelph, ON, Canada, 27–31 May 2024. [Google Scholar]
- Petsiuk, V.; Das, A.; Saenko, K. RISE: Randomized input sampling for explanation of black-box models. In Proceedings of the British Machine Vision Conference, Northumbria University, Newcastle, UK, 3–6 September 2018. [Google Scholar]
- Tian, H.; Zheng, Y.; Jin, Z. Improved RetinaNet model for the application of small target detection in the aerial images. In IOP Conference Series: Earth and Environmental Science; IOP Publishing: Bristol, UK, 2020; Volume 585, p. 012142. [Google Scholar]
- Avazov, K.; Jamil, M.K.; Muminov, B.; Abdusalomov, A.B.; Cho, Y.I. Fire detection and notifcation method in ship areas using deep learning and computer vision approaches. Sensors 2023, 23, 7078. [Google Scholar] [CrossRef]
- Geng, X.; Su, Y.; Cao, X.; Li, H.; Liu, L. YOLOFM: An improved fire and smoke object detection algorithm based onYOLOv5n. Sci. Rep. 2024, 14, 4543. [Google Scholar] [CrossRef]
- Nvidia DeepStream Documentation. Available online: https://developer.nvidia.com/deepstream-sdk (accessed on 2 July 2024).
Dataset | Classes | Num Images | Source | Content |
---|---|---|---|---|
Foggia et al. [11] | Fire/Non-fire | 62,690 | Ground cameras’ video frames | Indoor/outdoor scenes |
Li et al. [12] | Fire/Non-fire | 50,000 | Ground cameras’ video frames and static images | Indoor/outdoor scenes |
Yar et al. [13] | Fire/Non-fire | 3804 | Ground cameras’ video frames | Outdoor scenes |
Dilshad et al. [14] | Fire/Non-fire | 47,124 | Ground cameras’ images | Foggy, low-light outdoor scenes |
Shamsoshoara et al. [15] | Fire/Smoke/Other | 39,375 | Air-borne visible spectrum/infrared cameras | Forest fires |
Yar et al. [16] | Fire/Non-fire | 6000 | Air/space-borne cameras | Haze, fog, night outdoor scenes |
Wang et al. [17] | Fire/Smoke/Other | 122,634 | Ground/air-borne cameras | Indoor/outdoor scenes |
Parameter Name | Type |
---|---|
Operating system | Ubuntu 22.04.3 LTS |
CPU | Intel Core 9 Gen i9-13900KF (5.8 GHz) |
GPU | RTX 4090 24 GB |
RAM | DDR5 64 GB (6000 MHz) |
Deployment environment | Python 3.10.2 |
Deep learning framework | PyTorch 2.1.2 |
Accelerated computing architecture | CUDA 12.1 |
YOLOv5 | YOLOv6 | YOLOv7 | YOLOv8 | YOLOv9 | YOLOv10 | YOLO-NAS | |
---|---|---|---|---|---|---|---|
Epochs | 100 | 100 | 100 | 100 | 100 | 100 | 100 |
Batch size | 64 | 44 | 32 | 64 | 16 | 32 | 64 |
Optim. alg. | SGD | SGD | SGD | SGD | SGD | SGD | Adam |
Model weights | yolov5m | yolov6m | yolov7 | yolov8m | yolov9-c | yolov10m | yolo_nas_s |
Num. params. (million) | 21.2 | 34.9 | 35.9 | 25.9 | 50.7 | 19.1 | 19 |
Train | Validation | Test | |
---|---|---|---|
Nr of images | 20,108 | 4757 | 2417 |
Fire instances | 19,864 | 4490 | 2353 |
Smoke instances | 18,017 | 4282 | 2249 |
Other instances | 10,648 | 2691 | 1332 |
Model | Fire | Smoke | Other | Speed (fps) | ||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
P | R | mAP @50 | mAP @50–95 | F1 | P | R | mAP @50 | mAP @50–95 | F1 | P | R | mAP @50 | mAP @50–95 | F1 | ||
YOLOv5 | 0.86 | 0.86 | 0.91 | 0.69 | 0.86 | 0.88 | 0.82 | 0.89 | 0.66 | 0.85 | 0.80 | 0.78 | 0.84 | 0.69 | 0.79 | 210 |
YOLOv6 | 0.84 | 0.86 | 0.91 | 0.64 | 0.85 | 0.89 | 0.81 | 0.91 | 0.63 | 0.84 | 0.78 | 0.73 | 0.81 | 0.56 | 0.75 | 215 |
YOLOv7 | 0.83 | 0.88 | 0.91 | 0.64 | 0.85 | 0.87 | 0.84 | 0.90 | 0.62 | 0.86 | 0.76 | 0.79 | 0.84 | 0.58 | 0.77 | 230 |
YOLOv8 | 0.86 | 0.85 | 0.91 | 0.69 | 0.85 | 0.88 | 0.84 | 0.91 | 0.68 | 0.85 | 0.82 | 0.77 | 0.85 | 0.63 | 0.79 | 230 |
YOLOv9 | 0.85 | 0.85 | 0.91 | 0.69 | 0.85 | 0.88 | 0.85 | 0.91 | 0.68 | 0.85 | 0.76 | 0.68 | 0.84 | 0.63 | 0.72 | 115 |
YOLOv10 | 0.86 | 0.85 | 0.91 | 0.69 | 0.85 | 0.89 | 0.83 | 0.91 | 0.68 | 0.86 | 0.82 | 0.74 | 0.84 | 0.63 | 0.78 | 210 |
YOLO-NAS | - | - | 0.84 | - | - | - | - | 0.80 | - | - | - | - | 0.70 | - | - | - |
YOLOv5 | YOLOv6 | YOLOv7 | YOLOv8 | YOLOv9 | YOLOv10 | YOLO-NAS | ||
---|---|---|---|---|---|---|---|---|
Confidence score > 0.5 | ||||||||
Number of detections (Average confidence score) | Fire | 13,025 (0.799) | 14,635 (0.730) | 14,761 (0.770) | 12,703 (0.743) | 12,874 (0.719) | 12,968 (0.769) | 15,414 (0.717) |
Smoke | 21,694 (0.792) | 22,145 (0.766) | 21,895 (0.758) | 20,948 (0.768) | 20,204 (0.759) | 20,733 (0.796) | 17,439 (0.722) | |
Confidence score > 0.8 | ||||||||
Number of detections (Average confidence score) | Fire | 8937 (0.882) | 5570 (0.885) | 7949 (0.883) | 4762 (0.865) | 2733 (0.876) | 6715 (0.885) | 4447 (0.891) |
Smoke | 13,033 (0.892) | 11,530 (0.890) | 10,303 (0.885) | 11,954 (0.880) | 10,227 (0.887) | 13,389 (0.895) | 5405 (0.885) | |
Speed (fps) | 207 | 214 | 179 | 197 | 116 | 210 | 31 |
YOLOv5 | YOLOv6 | YOLOv7 | YOLOv8 | YOLOv9 | YOLOv10 | YOLO-NAS | ||
---|---|---|---|---|---|---|---|---|
Confidence score > 0.5 | ||||||||
Number of detections (Average confidence score) | Fire | 7053 (0.836) | 7472 (0.772) | 7533 (0.796) | 7114 (0.786) | 6358 (0.778) | 6606 (0.835) | 5408 (0.781) |
Smoke | 6105 (0.744) | 5777 (0.673) | 6152 (0.693) | 4752 (0.722) | 5181 (0.721) | 4621 (0.743) | 3020 (0.720) | |
Confidence score > 0.8 | ||||||||
Number of detections (Average confidence score) | Fire | 5540 (0.891) | 4216 (0.889) | 4981 (0.889) | 4762 (0.865) | 3716 (0.883) | 5086 (0.896) | 2968 (0.890) |
Smoke | 2699 (0.889) | 1705 (0.887) | 1867 (0.881) | 1959 (0.880) | 2107 (0.890) | 2214 (0.893) | 940 (0.892) | |
Speed (fps) | 180 | 219 | 245 | 240 | 114 | 216 | 31 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Catargiu, C.; Cleju, N.; Ciocoiu, I.B. A Comparative Performance Evaluation of YOLO-Type Detectors on a New Open Fire and Smoke Dataset. Sensors 2024, 24, 5597. https://doi.org/10.3390/s24175597
Catargiu C, Cleju N, Ciocoiu IB. A Comparative Performance Evaluation of YOLO-Type Detectors on a New Open Fire and Smoke Dataset. Sensors. 2024; 24(17):5597. https://doi.org/10.3390/s24175597
Chicago/Turabian StyleCatargiu, Constantin, Nicolae Cleju, and Iulian B. Ciocoiu. 2024. "A Comparative Performance Evaluation of YOLO-Type Detectors on a New Open Fire and Smoke Dataset" Sensors 24, no. 17: 5597. https://doi.org/10.3390/s24175597