Ricerca CERCA

DAUIN - GR-23 - VANDAL - Visual and Multimodal Applied Learning Lab

Deep Learning Techniques for Image Generation from Music

azienda Thesis in external company    

Reference persons TATIANA TOMMASI

External reference persons Giuseppe Rizzo (giuseppe.rizzo@linksfoundation.com), Angelica Urbanelli (angelica.urbanelli@linksfoundation.com)

Research Groups DAUIN - GR-23 - VANDAL - Visual and Multimodal Applied Learning Lab

Description Generative Adversarial Networks are a widespread technology that has been investigated for a lot of different tasks and applications, such as dataset generation for computer vision tasks or text to image generation.
Generative Adversarial Networks are a collection of Deep Neural Networks that are used in Computer Vision applications to generate images in a specific domain (pictures, faces, paintings), so that they end up producing images different from the ones seen at training time, though very realistic.
In this thesis, the applicant will study the state of the art in Generative Adversarial Networks and will analyse the existing tools and datasets for image generation, possibly extending them. The goal is to develop a system that will be able to generate artistic images starting from classical music tracks in a controlled way, that is, generating pictures able to represent the music's sentiment.

Required skills deep neural network models, python programming

Deadline 05/06/2023      PROPONI LA TUA CANDIDATURA

© Politecnico di Torino
Corso Duca degli Abruzzi, 24 - 10129 Torino, ITALY