Table 8 Comparisons of the computational cost of different methods on the Chikusei dataset with an NVIDIA RTX A6000 GPU.

From: Cross-range self-attention single hyperspectral image super-resolution method based on U-Net architecture

Model

d

Params. (\(\times 10^6\))

FLOPs (\(\times 10^9\))

TT (per epoch)

IT(per image)

GDRRN

4

0.4424

0.4152

16m6s

0.2870s

Deep_hs_Prior

4

2.2611

1.3167

\(\backslash\)

\(\backslash\)

SSPSR

4

13.5644

43.4842

24m15s

1.2130s

PDE_Net

4

2.2948

82.1785

27m18s

0.4259s

PUDPN

4

64.3712

95.2633

4h48m

2.0833s

CST

4

11.1131

5.3229

1h57m

1.6204s

Cs_Unet (Ours)

4

15.2633

64.3712

33m

3.3086s

Back to article page