Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : May 27, 2020 - May 30, 2020
This research aims to develop a monocular depth estimation method for robot navigation．In the robot navigation，object detection is important for local path planning．Recently most of autonomous driving cars and autonomous robots use LiDARs or stereo cameras for object detection．However，one of the problems in using a LiDAR for navigation is the cost of the sensor．Comparing to LiDARs，using stereo cameras is much less costly，but it needs to be calibrated properly to get accurate depth data． Therefore in this research we use a monocular camera and adopt a depth estimation using a deep neural network．We evaluate the performance of monocular camera-based depth estimation by comparing with the data from a stereo camera．
