Table 1.

Image processing time, network prediction time and prediction error of the present MLP using sample images with different resolutions. The results are compared with the performance of a DCNN-based approach and an inverse method

	Image processing time (ms)	Prediction time (ms)	Prediction error (%)
MLP (42 pixels, 20 nodes)	0.25	0.19	12
MLP (42 pixels, 60 nodes)	0.47	0.2	13
MLP (68 pixels, 20 nodes)	0.32	0.19	11
MLP (68 pixels, 60 nodes)	0.54	0.2	6
MLP (138 pixels, 20 nodes)	0.45	0.19	6
MLP (138 pixels, 60 nodes)	0.76	0.2	2
DCNN (138 pixels)	0.44	54.2	14
Inverse method (138 pixels, 60 nodes)	0.76	18.7	2