Table 2 Comprehensive performance metrics of the proposed XTinyHAR model evaluated on UTD-MHAD and MM-Fit datasets.
Metric | UTD-MHAD | MM-Fit |
|---|---|---|
Train accuracy (%) | 99.01 | 98.79 |
Validation accuracy (%) | 98.64 | 98.42 |
Test accuracy (%) | 98.71 | 98.55 |
Precision (%) | 98.72 | 98.56 |
Recall (%) | 98.71 | 98.55 |
F1-score (%) | 98.71 | 98.55 |
Cohen’s Kappa | 0.985 | 0.983 |
Model size (MB) | 2.45 | 2.45 |
Memory footprint (MB) | 7.2 | 7.1 |
Inference time CPU (ms) | 3.1 | 3.0 |
Inference time GPU (ms) | 1.2 | 1.1 |
Floating-point operations per second (FLOPs) (M) | 11.3 | 11.3 |
