top of page

Obj Following

Object Detection:
-
ROS Melodic cooperating Gazebo 9 and Rviz in Ubuntu 18.04
-
Assembled a RGB-D camera, perpendicular to the ground
-
SSD-based detection with MobileNet v2 as the backbone convolutional network
-
FasterRCNN-ResNet50
-
Environment: Gazebo Empty World
-
Objects: a white bowl, a coca can, a mobile robot, and a red ball




I set different flight heights: 1m, 1.5m, 2m, 2.5m, and the detected object for test is the same model “walking person” in Gazebo. ​
​
Problems:
-
Camera resolution is not high enough
-
All in simulator
-
Not familiar with the communications between different software
​


bottom of page