Fast, Accurate, and Lightweight Super-Resolution with Cascading Residual Network

ECCV 2018  ·  Namhyuk Ahn, Byungkon Kang, Kyung-Ah Sohn ·

In recent years, deep learning methods have been successfully applied to single-image super-resolution tasks. Despite their great performances, deep learning methods cannot be easily applied to real-world applications due to the requirement of heavy computation... In this paper, we address this issue by proposing an accurate and lightweight deep network for image super-resolution. In detail, we design an architecture that implements a cascading mechanism upon a residual network. We also present variant models of the proposed cascading residual network to further improve efficiency. Our extensive experiments show that even with much fewer parameters and operations, our models achieve performance comparable to that of state-of-the-art methods. read more

PDF Abstract ECCV 2018 PDF ECCV 2018 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Image Super-Resolution BSD100 - 2x upscaling CARN [[Ahn et al.2018]] PSNR 32.09 # 9
Image Super-Resolution BSD100 - 4x upscaling CARN PSNR 27.58 # 19
SSIM 0.7349 # 26
Image Super-Resolution Set14 - 2x upscaling CARN-M [[Ahn et al.2018]] PSNR 33.26 # 13
Image Super-Resolution Set14 - 2x upscaling CARN [[Ahn et al.2018]] PSNR 33.52 # 12
Image Super-Resolution Set14 - 4x upscaling CARN PSNR 28.60 # 20
SSIM 0.7806 # 26
Image Super-Resolution Set5 - 2x upscaling CARN [[Ahn et al.2018]] PSNR 37.76 # 13
Image Super-Resolution Set5 - 4x upscaling CARN PSNR 32.13 # 22
SSIM 0.8937 # 26
Image Super-Resolution Urban100 - 4x upscaling CARN PSNR 26.07 # 19
SSIM 0.7837 # 20

Methods


No methods listed for this paper. Add relevant methods here