Portal de Periódicos da CAPES

The Evolutionary Deep Learning based on Deep Convolutional Neural Network for the Anime Storyboard Recognition

2017; Springer Nature; Linguagem: Inglês

10.1007/978-3-319-62410-5_34

2194-5357

Saya Fujino, Taichi Hatanaka, Naoki Mori, Keinosuke Matsumoto,

Human Motion and Animation

Recently, the researches of image recognition have been developed remarkably by means of the deep learning. In this study, we focused on the anime storyboards and applied deep convolutional neural networks (DCNNs) to those data. There exists one problem that it takes a lot of effort to tune DCNN hyperparameters. To solve this problem, we propose a novel method called evolutionary the deep learning (evoDL) by means of genetic algorithms (GAs). The effectiveness of evoDL is confirmed by computer simulations taking a real anime storyboard recognition problem as an example.