Open AccessProceedings Article
Focal Frequency Loss for Image Reconstruction and Synthesis
Liming Jiang,Bo Dai,Wayne Wu,Chen Change Loy +3 more
- 01 Jan 2021
pp 13919-13929
TL;DR: In this article, the authors propose a novel focal frequency loss, which allows a model to adaptively focus on frequency components that are hard to synthesize by down-weighting the easy ones.
read more
Abstract: Image reconstruction and synthesis have witnessed remarkable progress thanks to the development of generative models. Nonetheless, gaps could still exist between the real and generated images, especially in the frequency domain. In this study, we show that narrowing gaps in the frequency domain can ameliorate image reconstruction and synthesis quality further. We propose a novel focal frequency loss, which allows a model to adaptively focus on frequency components that are hard to synthesize by down-weighting the easy ones. This objective function is complementary to existing spatial losses, offering great impedance against the loss of important frequency information due to the inherent bias of neural networks. We demonstrate the versatility and effectiveness of focal frequency loss to improve popular models, such as VAE, pix2pix, and SPADE, in both perceptual quality and quantitative performance. We further show its potential on StyleGAN2.
read more
Chat with Paper
AI Agents for this Paper
Find similar papers on Google Scholar, PubMed and Arxiv
Write a critical review of this paper
Analyze citations of this paper to find unaddressed research gaps
Citations
MAXIM: Multi-Axis MLP for Image Processing
01 Jun 2022
TL;DR: MAXIM as discussed by the authors uses a UNet-shaped hierarchical structure and supports long-range interactions enabled by spatially-gated MLPs, which can serve as an efficient and flexible general-purpose vision backbone for image processing tasks.
311
Camouflaged Object Detection with Feature Decomposition and Edge Reconstruction
Chunming He,Kaixuan Li,Yachao Zhang,Longxiang Tang,Yulun Zhang,Zhenhua Guo,Xiu Li +6 more
- 01 Jun 2023
TL;DR: By learning the auxiliary task in conjunction with the COD task, the FEDER model can generate precise prediction maps with accurate object boundaries and significantly outperforms state-of-the-art methods with cheaper computational and memory costs.
178
Deep Fourier-Based Exposure Correction Network with Spatial-Frequency Interaction
Jie Huang,Yajing Liu,Fengmei Zhao,Jinghao Zhang,Yukun Huang,Man Zhou,Zhiwei Xiong +6 more
- 01 Jan 2022
TL;DR: Huang et al. as discussed by the authors proposed a deep Fourier-based exposure correction network (FECNet) consisting of an amplitude sub-network and a phase subnetwork to progressively reconstruct the representation of lightness and structure components.
109
HDNet: High-resolution Dual-domain Learning for Spectral Compressive Imaging
01 Jun 2022
TL;DR: In this article , a high-resolution dual-domain learning network (HDNet) is proposed for hyperspectral image reconstruction, which combines spatial-spectral attention and frequency domain learning.
Single-View View Synthesis in the Wild with Learned Adaptive Multiplane Images
Ruicheng Wang,Jiaolong Yang +1 more
- 24 May 2022
TL;DR: This paper designs a network structure that consists of two novel modules, one for plane depth adjustment and another for depth-aware color prediction, and proposes a new method based on the multiplane image (MPI) representation for synthesizing novel views for in-the-wild photographs.
References
•Proceedings Article
CNNpack: packing convolutional neural networks in the frequency domain
Yunhe Wang,Chang Xu,Shan You,Dacheng Tao,Chao Xu +4 more
- 05 Dec 2016
TL;DR: In this article, the authors proposed to decompose convolutional filters as common parts (i.e., cluster centers) shared by other similar filters and their individual private parts (e.g., individual residuals).
159
Unsupervised Real-world Image Super Resolution via Domain-distance Aware Training
Yunxuan Wei,Shuhang Gu,Yawei Li,Radu Timofte,Longcun Jin,Hengjie Song +5 more
- 20 Jun 2021
TL;DR: Gu et al. as mentioned in this paper proposed a domain-gap aware super-resolution (DASR) approach for unsupervised real-world image SR, which takes advantage of real data in the target domain while domain-distance weighted supervision brings forward the more rational use of labeled source domain data.
TSIT: A Simple and Versatile Framework for Image-to-Image Translation
Liming Jiang,Changxu Zhang,Mingyang Huang,Chunxiao Liu,Jianping Shi,Chen Change Loy +5 more
- 23 Aug 2020
TL;DR: TSIT as mentioned in this paper proposes a coarse-to-fine generative model to capture semantic structure information and style representation by the network, allowing it to scale to various tasks in both unsupervised and supervised settings.
109
•Proceedings Article
Faster Neural Networks Straight from JPEG.
Lionel Gueguen,Alex Sergeev,Rosanne Liu,Jason Yosinski +3 more
- 01 Jan 2018
92
•Posted Content
Frequency Domain Image Translation: More Photo-realistic, Better Identity-preserving
TL;DR: A novel frequency domain image translation (FDIT) framework, exploiting frequency information for enhancing the image generation process, and effectively preserves the identity of the source image, and produces photo-realistic images.