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

Mustafa Ayazoglu; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2021, pp. 2472-2479

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 introduced in DeepRoots for Image Classification. We succesfully applied root modules to SISR problem, further more to make the model uint8 quantizaiton robust we used Clipped ReLU at the last layer of the network and achieved great balance between reconstruction quality and runtime. Further more although the proposed network contains 30x fewer parameters than VDSR it's performance surpasses it on Div2K validation set. The network proved itself by winning Mobile AI 2021 Real-Time Single Image Super Resolution Challenge.