Overparametrization of HyperNetworks at Fixed FLOP-Count Enables Fast Neural Image Enhancement
This addresses the need for efficient image enhancement on mobile devices, offering a novel method to decouple FLOPs from parameters, though it is incremental in applying HyperNetworks to a specific domain.
The paper tackled the problem of reducing computational cost (FLOPs) for neural image enhancement on mobile devices while maintaining high performance, achieving state-of-the-art results in SSIM and MS-SSIM on the ZRR dataset with over 10x reduced FLOP-count.
Deep convolutional neural networks can enhance images taken with small mobile camera sensors and excel at tasks like demoisaicing, denoising and super-resolution. However, for practical use on mobile devices these networks often require too many FLOPs and reducing the FLOPs of a convolution layer, also reduces its parameter count. This is problematic in view of the recent finding that heavily over-parameterized neural networks are often the ones that generalize best. In this paper we propose to use HyperNetworks to break the fixed ratio of FLOPs to parameters of standard convolutions. This allows us to exceed previous state-of-the-art architectures in SSIM and MS-SSIM on the Zurich RAW- to-DSLR (ZRR) data-set at > 10x reduced FLOP-count. On ZRR we further observe generalization curves consistent with 'double-descent' behavior at fixed FLOP-count, in the large image limit. Finally we demonstrate the same technique can be applied to an existing network (VDN) to reduce its computational cost while maintaining fidelity on the Smartphone Image Denoising Dataset (SIDD). Code for key functions is given in the appendix.