This research has immediate applications to driverless cars. Much of the machine learning used to bring autonomous technology to its current level has dealt with two dimensional images—still photos. A computer shown several million photos of a stop sign will eventually come to recognize stop signs in the real world and in real time.
But by utilizing video clips that run for several seconds, the U-M system can study the first half of the snippet to make its predictions, and then verify the accuracy with the second half.
“If a pedestrian is playing with their phone, you know they’re distracted,” said Ram Vasudevan, a U-M assistant professor of mechanical engineering. “Their pose and where they’re looking is telling you a lot about their level of attentiveness. It’s also telling you a lot about what they’re capable of doing next.”
The research was conducted out of the U-M Ford Center for Autonomous Vehicles (FCAV) by Xiaoxiao Du, a research engineer in FCAV, Matthew Johnson-Roberson, an associate professor of naval architecture and marine engineering, and Vasudevan.
https://fcav.engin.umich.edu
Read the paper: "Bio-LSTM: A Biomechanically Inspired Recurrent Neural Network for 3D Pedestrian Pose and Gait Prediction" in IEEE Robotics and Automation Letters, 2019:
https://doi.org/10.1109/LRA.2019.2895266
This video was produced by the FCAV lab, which acknowledges one of its former research engineers, Charles Barto, for his help in making this video, and also thanks Wonhui Kim and the rest of FCAV lab members who helped providing the PedX dataset used in this video.
https://www.engin.umich.edu
Социальные закладки