This project extends a previous implementation of tensorflow object detection in ROS. Using a depth camera, pixel coordinates are reprojected into 3D and published to /tf.
Tensorflow and ROS
This guide targets Ubuntu 16.04 and ROS Kinetic
Start the rosnode of your depth camera. For example,
roslaunch realsense2_camera rs_camera.launch
Then, run
rosrun ros_object_detection_3d detect_ros_and_find_depth.py