This repo introduces how to integrate Tensorflow framework into ROS with object detection API.
And through this repo, you can realize mnist, object recognition, and object detection respectively.
sudo apt-get install protobuf-compiler python-pil python-lxml
sudo pip install jupyter
sudo pip install matplotlib
cd [CATKIN_WS]/src
git clone https://github.com/cong/ros_tensorflow.git
Build ROS package by
cd [CATKIN_WS]
catkin_make
PS: Python doesn't seem to need to be compiled.
# First, open a terminal, execute
roscore
# Second, Open another terminal, then execute
# Please install "usb_cam" node before you execute blow
roslaunch usb_cam usb_cam-test.launch
# Third, open another terminal, then execute
roslaunch ros_tensorflow ros_tensorflow_mnist.launch
# You can echo a topic to receive the string message.
rostopic echo /result_ripe
# Third, open another terminal, then execute
roslaunch ros_tensorflow ros_tensorflow_classify.launch
# You can echo a topic to receive the string message.
rostopic echo /result_ripe
# Third, open another terminal, then execute
roslaunch ros_tensorflow ros_tensorflow_detect.launch
# You can through "image_view" node to receive images detected.
rosrun image_view image_view image:=/result_ripe
Publish a topic : /result_ripe
Receive an image : usb_cam/image_raw