Neural Architecture Search for Image Super-Resolution Using Densely\n Constructed Search Space: DeCoNAS

Résumé

The recent progress of deep convolutional neural networks has enabled great\nsuccess in single image super-resolution (SISR) and many other vision tasks.\nTheir performances are also being increased by deepening the networks and\ndeveloping more sophisticated network structures. However, finding an optimal\nstructure for the given problem is a difficult task, even for human experts.\nFor this reason, neural architecture search (NAS) methods have been introduced,\nwhich automate the procedure of constructing the structures. In this paper, we\nexpand the NAS to the super-resolution domain and find a lightweight densely\nconnected network named DeCoNASNet. We use a hierarchical search strategy to\nfind the best connection with local and global features. In this process, we\ndefine a complexity-based penalty for solving image super-resolution, which can\nbe considered a multi-objective problem. Experiments show that our DeCoNASNet\noutperforms the state-of-the-art lightweight super-resolution networks designed\nby handcraft methods and existing NAS-based design.\n

Citer ce document

Accès au document

Ce document est hébergé sur son dépôt institutionnel d'origine.

Auteur(s)

Statistiques

Consultations : 1

Téléchargements : 0