Neural architecture search for deep image prior
| Autor: | Hailin Jin, Andrew Gilbert, Kary Ho, John Collomosse |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Structure (mathematical logic)
Computer science business.industry General Engineering Inpainting Binary number 020207 software engineering Pattern recognition 02 engineering and technology Computer Graphics and Computer-Aided Design Image (mathematics) Human-Computer Interaction Range (mathematics) 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Artificial intelligence Architecture Image denoising business Design space |
| Zdroj: | Computers & Graphics. 98:188-196 |
| ISSN: | 0097-8493 |
| DOI: | 10.1016/j.cag.2021.05.013 |
| Popis: | We present a neural architecture search (NAS) technique to enhance image denoising, inpainting, and super-resolution tasks 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. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect