Table 16.

Evaluation of RPi using TFLite and quantization-aware training with 8-bit on the Urban Sound dataset.

Models	Inference Time (ms)	${\mathit{F}}_1$ Micro	${\mathit{F}}_1$ Macro	${\mathit{F}}_1$ Weighted
ANN	0.121 ± 0.009	0.942	0.950	0.944
CNN-1D	18.856 ± 0.171	0.951	0.960	0.951