object_detection-

Repo for object detection/localization in ROS with Intel Realsense camera

Setup

Requirements

1) Install ROS melodic

Follow the guide here to install ROS melodic
paste this into the terminal to install catkin build tools :

sudo apt-get install python-catkin-tools

2) Install Torch (used to run neural network inference)

Go to the Nvidia Website here and download the PyTorch v1.8.0 .whl file
extract the folder in the home directory and rename it to torch (so the compiler can find it)
after that the torch folder will be in your home directory (see below)

follow the other instructions made in the installation section by Nvidia :

wget https://nvidia.box.com/shared/static/p57jwntv436lfrd78inwl7iml6p13fzh.whl -O torch-1.8.0-cp36-cp36m-linux_aarch64.whl
sudo apt-get install python3-pip libopenblas-base libopenmpi-dev 
pip3 install Cython
pip3 install numpy torch-1.8.0-cp36-cp36m-linux_aarch64.whl

After that torch should be installed

Install

1) Create a folder called `CATKIN_FS/src/` in your home directory :

mkdir -p CATKIN_FS/src/

2) Clone the repo

git clone https://github.com/Cedric-Perauer/object_detection-.git

3) Move files from the object_detection- folder to the CATKIN_FS/src/ folder :

mv object_detection-/* CATKIN_FS/src/

4) go into the CATKIN_FS directory and build the project with catkin build :

cd CATKIN_FS/

catkin build

5) generate the .engine file :

TODO

6) Change the Line 65 in camera_node/src/yolov5.cpp to point to the .engine file location (replace user name with your username) :

 engine_name = "/home/<user_name>/CATKIN_FS/src/camera_node/src/yolov5s.engine";

do a similar thing in the file camera_node/src/camera_node.cpp (line 23)

std::string paramsFilePath = "/home/<user_name>/CATKIN_FS/camera_node/src/params.txt";

6) go into the CATKIN_FS directory and build the project with catkin build :

cd CATKIN_FS/

catkin build

=> When the build is successful you can download the rosbag file using this link

go into the downloads directory where the .bag file is located and run it with :

rosbag play -l camera.bag

start the camera_node wit :

rosrun camera_node camera_node

then you should see an image window openig that looks like this :

Potential Issues Fix :

If you run into an issue regarding boost (python_boost) or opencv follow these steps :

1) install OpenCv3 version (paste into terminal) :

sudo apt install libopencv-dev=3.2.0+dfsg-4ubuntu0.1

2) Change the 2 files below :

Commenting row 11 12 in the CMakeLists.txt in the folder vision_opencv/cv_bridge/

if(PYTHONLIBS_VERSION_STRING VERSION_LESS "3.8")
    # Debian Buster
    #find_package(Boost REQUIRED python37)
    #else()
    # Ubuntu Focal
    find_package(Boost REQUIRED python)
  endif()

And adding 0 at the #define NUMPY_IMPORT_ARRAY_RETVAL in the file /usr/include/python2.7/numpy/__multiarray_api.h

About