New AI Research From China Proposes Meta-Transformer: A Unified AI Framework for Multimodal Learning

New AI Research From China Proposes Meta-Transformer: A Unified AI Framework for Multimodal Learning

The human brain, considered the paradigm of neural network theories, simultaneously processes information from various sensory inputs, such as visual, auditory and tactile signals. Furthermore, understanding from one source might aid knowledge from another. However, due to the huge modality gap in deep learning, building a unified network that can process various forms of input…

ByteDance AI Research proposes a new self-supervised learning framework to create high-quality stylized 3D avatars with a mix of continuous and discrete parameters

ByteDance AI Research proposes a new self-supervised learning framework to create high-quality stylized 3D avatars with a mix of continuous and discrete parameters

A key entry point into the digital world, which is more prevalent in modern life for socializing, shopping, playing games and other activities, is a visually appealing and animated 3D avatar. A decent avatar should be attractive and customized to fit the user’s appearance. Many popular avatar systems, such as Zepeto1 and ReadyPlayer2, use animated…

New AI research proposes a simple but effective structure-based encoder for learning the representation of proteins based on their 3D structures

New AI research proposes a simple but effective structure-based encoder for learning the representation of proteins based on their 3D structures

Proteins, the cell’s energy, are involved in various applications, including materials and treatments. They consist of a chain of amino acids that folds into a certain shape. A significant number of new protein sequences have recently been discovered thanks to the development of low-cost sequencing technology. Accurate and effective in silico protein function annotation methods…