Table 14 Comparison of accuracy vs. loss analysis (NASA and fire videos datasets).
From: Real time fire and smoke detection using vision transformers and spatiotemporal learning
Model | Dataset | Training accuracy (%) | Training loss | Validation accuracy (%) | Validation loss | Test accuracy (%) | Test loss |
|---|---|---|---|---|---|---|---|
Proposed hybrid model | NASA | 99.2 | 0.05 | 98.5 | 0.18 | 98.3 | 0.21 |
Fire videos | 99.1 | 0.05 | 98.4 | 0.19 | 98.3 | 0.22 | |
ResNet50 | NASA | 98.5 | 0.06 | 95.5 | 0.30 | 95.2 | 0.33 |
Fire videos | 98.0 | 0.07 | 94.8 | 0.32 | 94.3 | 0.36 | |
VGG16 | NASA | 97.8 | 0.07 | 93.2 | 0.35 | 92.8 | 0.38 |
Fire videos | 97.2 | 0.08 | 92.5 | 0.37 | 92.0 | 0.40 | |
LSTM | NASA | 96.0 | 0.08 | 92.0 | 0.40 | 91.6 | 0.45 |
Fire videos | 95.5 | 0.09 | 91.3 | 0.42 | 90.8 | 0.47 | |
3D-CNN | NASA | 96.8 | 0.07 | 93.5 | 0.36 | 93.0 | 0.40 |
Fire videos | 96.2 | 0.08 | 92.8 | 0.38 | 92.4 | 0.42 | |
Hybrid ResNet50 + LSTM | NASA | 98.2 | 0.06 | 96.0 | 0.27 | 95.8 | 0.30 |
Fire videos | 97.8 | 0.06 | 95.5 | 0.28 | 95.3 | 0.31 | |
Hybrid VGG16 + 3D-CNN | NASA | 98.0 | 0.06 | 95.8 | 0.29 | 95.5 | 0.32 |
Fire videos | 97.5 | 0.07 | 95.0 | 0.31 | 94.8 | 0.34 |
