arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2001.04776cs.CV
收藏

深度图像先验的神经结构搜索

Neural Architecture Search for Deep Image Prior

University of Surrey(萨里大学);Adobe Research(奥多比研究院)

Kary Ho, Andrew Gilbert, Hailin Jin, John Collomosse

摘要:

我们提出了一种神经架构搜索(NAS)技术,以提高最近提出的深度图像先验(DIP)下的无监督图像去噪,修复和超分辨率的性能。我们表明,进化搜索可以自动优化的编码器-解码器(E-D)的结构和DIP网络的元参数,这作为一个特定的内容之前,规范这些单一的图像复原任务。我们的二进制表示对非对称E-D网络的设计空间进行编码,该网络通常会在10-20代内收敛,使用500个种群规模产生特定于内容的DIP。经过优化的架构持续改进经典DIP的视觉质量,适用于各种摄影和艺术内容。

英文摘要:

We present a neural architecture search (NAS) technique to enhance the performance of unsupervised image de-noising, in-painting and super-resolution under the recently proposed Deep Image Prior (DIP). We show that evolutionary search can automatically optimize the encoder-decoder (E-D) structure and meta-parameters of the DIP network, which serves as a content-specific prior to regularize these single image restoration tasks. Our binary representation encodes the design space for an asymmetric E-D network that typically converges to yield a content-specific DIP within 10-20 generations using a population size of 500. The optimized architectures consistently improve upon the visual quality of classical DIP for a diverse range of photographic and artistic content.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑