Yingtao Tian - Computational Creativity in Abstract Art: Exploring 2D and 3D Art with Evolutionary Algorithms
Host
Stefanie Mueller
CSAIL MIT
Abstract:
Computational creativity has a significant role in modern-era abstract arts, where an ever-lasting quest is to assist artists in creating high quality, abstract art with computational approaches. However, in this context, two recent challenges have emerged: archiving high-quality abstract painting creation comparable to those produced by recent promising deep learning-based approaches, and extending beyond two-dimensional art to address abstract 3D art with high-quality and controllability. These challenges stem from the difficulty in defining gradient-based models and the varying interpretations of abstract arts.
In this talk, we will present our research works aimed at advancing computational creativity in the direction of abstract art creation, with an emphasis on both 2D and 3D abstract art forms. We will discuss two key studies: "Modern Evolution Strategies for Creativity: Fitting Concrete Images and Abstract Concepts" and "Evolving Three-Dimensional (3D) Abstract Art: Fitting Concepts by Language." Both investigations utilize modern evolutionary algorithms to tackle the challenges of abstract art creation, where a well-defined gradient-based optimization process is hard to define.
In this first study, we explore the use of modern evolutionary algorithms for generating two-dimensional abstract art that conforms to textual or visual prompts. In the second study, we propose to bridge modern evolution strategies and 3D rendering through customizable parameterization to produce 3D scenes. These scenes, when rendered into films, follow artists' textual specifications. While our works focus on a limited set of abstract art forms, we hope they could offer a fresh perspective for artists to easily express creative ideas for abstract art, and serve as an inspiration for future works in the field.
Bio:
Yingtao Tian is a Research Scientist at Google Brain Tokyo. He obtained his PhD in Computer Science at Stony Brook University and B.S. in Computer Science and Technology at Fudan University. His research interests lie in generative models and representation learning, as well as their applications in image generation, natural language processing, knowledge base modeling, social network modeling, bioinformatics.
In addition to his core research interests, he is passionate about evolution strategies and the interdisciplinary study of machine intelligence and creativity within the context of humanities research. His work aims to advance the understanding and development of computational creativity and its impact on various artistic domains.
The talk will also be streamed over Zoom: https://mit.zoom.us/j/93441815357.
Computational creativity has a significant role in modern-era abstract arts, where an ever-lasting quest is to assist artists in creating high quality, abstract art with computational approaches. However, in this context, two recent challenges have emerged: archiving high-quality abstract painting creation comparable to those produced by recent promising deep learning-based approaches, and extending beyond two-dimensional art to address abstract 3D art with high-quality and controllability. These challenges stem from the difficulty in defining gradient-based models and the varying interpretations of abstract arts.
In this talk, we will present our research works aimed at advancing computational creativity in the direction of abstract art creation, with an emphasis on both 2D and 3D abstract art forms. We will discuss two key studies: "Modern Evolution Strategies for Creativity: Fitting Concrete Images and Abstract Concepts" and "Evolving Three-Dimensional (3D) Abstract Art: Fitting Concepts by Language." Both investigations utilize modern evolutionary algorithms to tackle the challenges of abstract art creation, where a well-defined gradient-based optimization process is hard to define.
In this first study, we explore the use of modern evolutionary algorithms for generating two-dimensional abstract art that conforms to textual or visual prompts. In the second study, we propose to bridge modern evolution strategies and 3D rendering through customizable parameterization to produce 3D scenes. These scenes, when rendered into films, follow artists' textual specifications. While our works focus on a limited set of abstract art forms, we hope they could offer a fresh perspective for artists to easily express creative ideas for abstract art, and serve as an inspiration for future works in the field.
Bio:
Yingtao Tian is a Research Scientist at Google Brain Tokyo. He obtained his PhD in Computer Science at Stony Brook University and B.S. in Computer Science and Technology at Fudan University. His research interests lie in generative models and representation learning, as well as their applications in image generation, natural language processing, knowledge base modeling, social network modeling, bioinformatics.
In addition to his core research interests, he is passionate about evolution strategies and the interdisciplinary study of machine intelligence and creativity within the context of humanities research. His work aims to advance the understanding and development of computational creativity and its impact on various artistic domains.
The talk will also be streamed over Zoom: https://mit.zoom.us/j/93441815357.