Table 5 Performance comparison of lightweight SR methods on the RK3588NPU. Tested on Self-built datasets. Inference time is measured based on the output image size of 1280\(\times\)720. The best and second-best performances are highlighted in Italic and bold, respectively.
From: Reparameterizable large kernel attention networks for infrared image super-resolution
REPLKA Module | Scale | Self-built NIQE/PIQE | time consuming(ms) |
|---|---|---|---|
ESPCN-uint8 | \(\times\)4 | 5.86/5.68 | 44.24 |
FSRCNN-uint8 | 5.58/5.63 | 53.32 | |
IMDN-RTC-uint8 | 5.24/4.99 | 51.67 | |
ECBSR-uint8 | 4.67/4.52 | 37.22 | |
REPLKASR-uint8 | 4.42/4.48 | 37.59 |
