Pioneering new domains in pattern recognition and image sensing
Deep learning is becoming indispensable in image pattern recognition. Our laboratory uses existing deep learning models, promotes research to further improve recognition accuracy and realize highly human-compatible recognition systems, such as new architecture and learning methods, and attempts to visualize and understand the internals of these . In addition, we pursue new image sensing methods, including image measurement, recognition and generation, with the aim of pioneering these domains.
Human movement analysis / behavior recognition technology
Our laboratory has acquired expressions for modeling human form and motion with high accuracy and efficiency, and advanced research on human recognition by machine learning. We are promoting research on robust human detection and tracking from images, posture estimation, motion analysis / prediction technology, and the eclectic application of these.
Sports video analysis
In sports, quantitative play analysis from images is important in improving the level of competition and motivation of athletes, supporting coaching, and providing new broadcast video content. Our laboratory is conducting research on methods and systems for sport image analysis which can be practically utilized in the field of various sports.
To date, robots have performed various services based on careful instructions. In our laboratory, we are conducting R&D on an intelligent robot that behaves appropriately by observing the situation and people, using real-time human behavior recognition, object/environment recognition technology and past action logs to obtain various forms of “awareness”.
Real world sensing
Image sensing technology is expected to be utilized in various situations in the real world. Our laboratory aims to utilize new image sensing technology in various fields such as automobiles and medical care.