Extremely Lightweight Quantization Robust Real-Time Single-Image Super Resolution for Mobile Devices

TL;DR

This paper introduces XLSR, an extremely lightweight, quantization-robust super-resolution network designed for mobile devices, achieving high performance with significantly fewer parameters and real-time processing capabilities.

Contribution

The paper presents a novel lightweight SISR model using root modules and quantization techniques, outperforming larger models like VDSR on mobile hardware.

Findings

Surpasses VDSR performance on Div2K validation set

Contains 30x fewer parameters than VDSR

Wins Mobile AI 2021 Real-Time SISR Challenge

Abstract

Single-Image Super Resolution (SISR) is a classical computer vision problem and it has been studied for over decades. With the recent success of deep learning methods, recent work on SISR focuses solutions with deep learning methodologies and achieves state-of-the-art results. However most of the state-of-the-art SISR methods contain millions of parameters and layers, which limits their practical applications. In this paper, we propose a hardware (Synaptics Dolphin NPU) limitation aware, extremely lightweight quantization robust real-time super resolution network (XLSR). The proposed model's building block is inspired from root modules for Image classification. We successfully applied root modules to SISR problem, further more to make the model uint8 quantization robust we used Clipped ReLU at the last layer of the network and achieved great balance between reconstruction quality and runtime. Furthermore, although the proposed network contains 30x fewer parameters than VDSR its performance surpasses it on Div2K validation set. The network proved itself by winning Mobile AI 2021 Real-Time Single Image Super Resolution Challenge.