Super-resolution guided knowledge distillation for low-resolution image classification

Hongyuan Chen, Yanting Pei, Hongwei Zhao, Yaping Huang

Abstract: Highlights•We propose a novel SRKD framework to solve the low-resolution classification problem.•Classification loss is introduced in GAN-based SR sub-network.•The features of HR images are used to guide the learning of the low-resolution network in the KD sub-network.•SRKD achieves state-of-the-art in the very low-resolution classification problem.