a ROS,ZED SSD,Tensorflow package

• A ros package for autonomous driving object detection.
• This performance is better than zed_ros_kerasssd_pkg

It contains modules of:

1. A tensorflow implemented SSD
2. Subscribe image info from ZED Camera ROS node, obtained the PointCloud information of Object bounding box center. calculate distance of object.
3. Publish a ROS message about the object information(object type/confidence/distance/)
4. ROS node name: new_detect_pkg
5. ROS message name: new_obj_info

• Ubuntu16.04
• CUDA 9.0
• CUDNN 7.0.5
• ROS Kinetic
• ZED Camera SDK 2.2.1
• zed-ros-wrapper-2.2.x
• Python version 3.5
• tensorflow-gpu (1.5.0rc1)
• cv-bridge (1.12.7)

• cv_bridge for python3

clone and unziped into your ROS workspace/src(suppose workspace names catkin_ws)

# for the first time
cd ~/catkin_ws 
catkin_make -DCATKIN_WHITELIST_PACKAGES=new_detect
cd src/detect_pkg/
chmod u+x bin/hello
cd ~/catkin_ws
. devel/setup.bash

# run it!!
roslaunch zed_wrapper zed.launch
# open a new terminal
rosrun new_detect hello

# check ROS Node statues
# open a new terminal
rostopic echo -p /new_obj_info

GTX GFORCE 1070: 7FPS