Generative Adversarial Networks for Game Graphic Asset Generation

2017.

Investigations on the generative adversarial network (GAN) models for learning graphic assets generation from retro game database and internet found images.

︎︎︎ ︎ NeurIPS Machine Learning for Creativity and Design, Art Gallery. (Link)

︎︎︎ ︎ AI Grant Fellow

︎ DCGAN generated game effect sprites.

︎Dataset: old school J-RPG FX animation sprites collected from game developer and fan art communities

DCGAN generated character sheet. ︎

︎Dataset: old school J-RPG character sprites collected from game developer and fan art communities.

︎DCGAN generated anime sky.
︎Dataset: collected images from Tumblr, tagged "anime" "sky"