Table 1 Inference time comparison of FIN, RH-M, and MH-PR Algorithms

From: Fourier Imager Network (FIN): A deep neural network for hologram reconstruction with superior external generalization

 

Number of trainable parameters

Inference time (s/mm2)

Parallelized inference time (s/mm2)

FIN (M = 3)

11.5 M

0.52

0.04

FIN (M = 4)

11.5 M

0.56

0.04

RH-M (M = 3)

14.1 M

4.84

1.96

RH-M (M = 4)

14.1 M

5.76

2.08

MH-PR (M = 2)

N/A

10.36

N/A

MH-PR (M = 3)

N/A

14.19

N/A

  1. The parallelized inference time is measured when the batch size is set to 20 for both FIN and RH-M
Back to article page