Fig 1.
Dataset folder structure.
Fig 2.
Genetic algorithm for hyperparameter optimization.
Fig 3.
XAI (Grad-Cam) architecture.
Fig 4.
Fire monitoring process.
Table 1.
Dataset distribution by class.
Fig 5.
Statistical analysis of indoor fire and smoke dataset labels.
Table 2.
The experimental environment settings.
Table 3.
Hyperparameters characteristics.
Table 4.
Proposed model performance by class.
Fig 6.
Improved YOLOv5s training results.
Fig 7.
Training progression curves: (a) Precision; (b) Recall; (c) F1; (d) P-R.
Fig 8.
Performance in varied settings: (a) Night environment; (b) Confined spaces; (c) Complex scenario; (d) Dark environment.
Table 5.
Comparative Analysis of SOTA object detection model performance metrics.
Fig 9.
Grad-CAM heatmap detection: (a) Fire; (b)Smoke.
Fig 10.
Fire hazard detection output.