Skip to main content

View full-text article in PMC

. Author manuscript; available in PMC: 2023 Jun 2.

Published in final edited form as: Med Image Anal. 2021 Apr 5;71:102048. doi: 10.1016/j.media.2021.102048

TABLE II.

Quantitative Result of Empirical Validation

Method	Metric	Data-Set
Method	Metric	DIC-C2DH-HeLa	Fluo-N2DH- GOWT1	Fluo-N2DH-SIM+	PhC-C2DH-U373
Original(CPU) [16], [36]	Time Cost (s/frame)	25.58	115.32	88.74	39.76
Original(CPU) [16], [36]	GPU Memory (GB)	5.92	5.92	5.92	5.92
GPU_H [21]	Time Cost (s/frame)	9.83	37.54	18.60	10.81
GPU_H [21]	GPU Memory (GB)	6.16	6.16	6.16	6.16
GPU_P [22]	Time Cost (s/frame)	-	-	-	-
GPU_P [22]	GPU Memory (GB)	overflow	overflow	overflow	overflow
Ours	Time Cost (s/frame)	2.74	11.24	11.72	5.64
Ours	GPU Memory (GB)	8.38	7.24	6.95	6.82

^*

5.92GB is the minimal GPU memory consumption for the deep learning based feature extraction.

^*

Due to the GPU overflow, the time-cost is not available(−)