In recent years, we have witnessed the spread of computer graphics techniques, used as a background map for movies and video games. Nevertheless, when creating 3D models with conventional computer graphics software, it is necessary for the user to manually change the placement and size. This requires expertise of computer graphics architecture and operations, which is time demanding. Applying Artificial Intelligence (AI) to games is currently an established research field. Starting from such premises, in this paper MONstEr (dEEp lEArNiNG GENErAtiON AssEt) a system for the automatic generation of virtual asset for videogames is presented. MONstEr exploits the principle of Deep Learning (DL) and in particular Generative models to automatically design new assets for videogames. The DL pipeline is the core of this system and it is based on a Deep Convolutional Generative Adversarial Network followed by Pixel2Mesh architecture for the 3D models generation. The approach was applied and tested on a newly collected dataset of images, “GameAssetDataset” which comprises characters representation extracted thanks a web crawler algorithm specifically developed for its acquisition. MONstEr expedites the implementation of solutions for new gamining environments, requiring only a small intervention in the 3D construction to insert the object in the game scene.

MONstEr: A Deep Learning-Based System for the Automatic Generation of Gaming Assets

Balloni E.;Paolanti M.;Frontoni E.;
2022-01-01

Abstract

In recent years, we have witnessed the spread of computer graphics techniques, used as a background map for movies and video games. Nevertheless, when creating 3D models with conventional computer graphics software, it is necessary for the user to manually change the placement and size. This requires expertise of computer graphics architecture and operations, which is time demanding. Applying Artificial Intelligence (AI) to games is currently an established research field. Starting from such premises, in this paper MONstEr (dEEp lEArNiNG GENErAtiON AssEt) a system for the automatic generation of virtual asset for videogames is presented. MONstEr exploits the principle of Deep Learning (DL) and in particular Generative models to automatically design new assets for videogames. The DL pipeline is the core of this system and it is based on a Deep Convolutional Generative Adversarial Network followed by Pixel2Mesh architecture for the 3D models generation. The approach was applied and tested on a newly collected dataset of images, “GameAssetDataset” which comprises characters representation extracted thanks a web crawler algorithm specifically developed for its acquisition. MONstEr expedites the implementation of solutions for new gamining environments, requiring only a small intervention in the 3D construction to insert the object in the game scene.
2022
978-3-031-13320-6
978-3-031-13321-3
File in questo prodotto:
File Dimensione Formato  
ICIAP_Game_DeepLearning_Workflow.pdf

solo utenti autorizzati

Tipologia: Documento in pre-print (manoscritto inviato all'editore, precedente alla peer review)
Licenza: Creative commons
Dimensione 2.4 MB
Formato Adobe PDF
2.4 MB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11393/305135
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 2
  • ???jsp.display-item.citation.isi??? 3
social impact