Fig 1.
A flowchart for methodology.
Fig 2.
Architecture of YOLOv10 [24].
Table 1.
Description and recommended ranges of key YOLO hyperparameters.
Fig 3.
GWO Strategy [36].
Fig 4.
Pedal Scent Marking and Sniffing Behaviors of Brown-Bears, (a) Pedal Marks in an Area.
(b-d) A brown-Bear Stretching to its Pedal Marks [39].
Fig 5.
The Flowchart of the Hybrid GWO-BBOA Optimization Process.
Fig 6.
Pseudocode of the proposed Hybrid GWO-BBOA algorithm.
Table 2.
Performance results of the YOLOv10s model on the Arson Detection Dataset.
Fig 7.
Evolution of training losses and evaluation metrics (precision, recall, mAP@0.5, mAP@0.5:0.95) over 100 epochs.
Table 3.
Performance results of YOLO v10 versions on arson dataset.
Table 4.
Parameters selected.
Table 5.
Performance results of YOLOv10s using different optimization algorithms on the arson detection dataset.
Fig 8.
Visual comparison of YOLOv10 arson detection results.
(A) Ground truth annotations for fire/arson events. (B) Predicted bounding boxes generated by YOLOv10 optimized with the proposed GWO-BBOA algorithm.
Table 6.
Pairwise T-test results comparing the proposed GWO-BBOA with PSO, GWO, and BBOA.
Table 7.
Presents a comparison between our study, Abbod et al. and Singh et al.
Fig 9.
Visualized results for a single arson scene, showing (A) the original image, (B) YOLO v10 bounding‑box, and (C) Grad‑CAM heatmap overlay.