Scientists developed a new way to create detailed, moving 3D objects from just one video. This breakthrough has potential to make it easier and more efficient to generate realistic animations for movies, games, or other applications.
Read Full Paper →Scientists created a new tool called SUB that allows them to test how well AI models can understand pictures by identifying specific features, like bird species or patterns. This benchmark could help make AI applications more reliable and trustworthy in fields like medicine, where accurate interpretation is crucial.
Read Paper →Scientists used a limited number of cameras to capture dynamic scenes like people repairing a bike or dancing, instead of needing hundreds of cameras. This new approach could make it easier and less expensive for people to reconstruct complex scenes from video footage, potentially opening up new possibilities for applications like virtual reality or surveillance.
Read Paper →Scientists created a computer system that can learn to interact with graphical user interfaces, such as clicking and typing on screens, by analyzing how humans perform these tasks. This new system achieved significantly better results than previous attempts at this task, which could lead to the development of more reliable and accurate "virtual assistants" in various areas.
Read Paper →We developed a new way to make 3D human models interact with their environment in a realistic way, by combining digital movement with physical rules that allow for safe and natural interactions. This approach has the potential to improve computer simulations and games by allowing characters to move around and interact with objects in a more believable and immersive way.
Read Paper →
