Artificial intelligence has made breakthrough progress in understanding human behavior. Researchers at Texas A&M University recently developed a new artificial intelligence system called "OmniPredict". The system has demonstrated unprecedented "mind-reading" capabilities - it can not only see human movements, but also predict human next intentions in real time by interpreting visual and environmental clues.
This research result marks a major leap in autonomous driving technology from "passive reaction" to "active intuition". Traditional autonomous driving systems can usually only identify the current location and movement trajectory of pedestrians, while OmniPredict introduces multi-modal large language model (MLLM) technology, giving it human-like reasoning capabilities. The system can keenly capture subtle signals such as changes in the pedestrian's posture, moments of hesitation, body orientation and even eye pressure, thereby inferring whether the pedestrian is preparing to cross the road, just waiting on the side of the road, or has other sudden behaviors.
The research team pointed out that the core advantage of OmniPredict is that it no longer just "looks" at various pixels, but tries to understand the "why" behind the behavior. By analyzing complex mixed input information, the model accurately categorizes human behavior into key categories such as crossing the road, sight occlusion, specific actions and gaze direction. In tests, OmniPredict demonstrated a prediction accuracy of up to 67%, a full 10 percentage points higher than the most advanced models currently on the market. What’s even more impressive is that even in complex scenarios where pedestrians are partially obscured or interact with the vehicle only through their eyes, the system still maintains extremely high judgment stability.
Project leader Dr. Srikanth Saripalli said that OmniPredict gives the machine a new kind of "Street Smarts". If self-driving cars can read the body language of passers-by and predict their next actions, just like human drivers, the safety of road traffic will make a qualitative leap. In addition to the field of autonomous driving, this technology that can interpret body language and psychological status is also expected to play a key role in high-risk scenarios such as military operations and emergency rescues in the future. By giving machines "intuition", it will completely change the model of human-machine collaboration.
Compiled from /ScitechDaily