tud-airlab / Ahold_product_detector

This repository implements a product detector for products in the Albert Heijn supermarket

Ahold Product Detector

This repository implements a product detector for products in the Albert Heijn supermarket. It is able to detect products using YoloV8 and track them using a Kalman Filter.

To run ROS with python, either a venv virtual environment or the conda or mamba package manager can be used. Specific instructions for these can be found below.

Venv Instructions

Install this ROS Package for Ubuntu

Before starting, make sure to have ROS Noetic installed: http://wiki.ros.org/noetic/Installation

After installation, initialize a workspace in order to build and run the ROS package. Do this by running the following commands:

mkdir YOUR_WORKSPACE_NAME
cd YOUR_WORKSPACE_NAME
mkdir src
cd src

We want to have the ROS package inside the 'src' directory, so now that we are in here we can clone the repository:

git clone [e-mail]:stijnla/Ahold_product_detector.git

Now return to your workspace directory, source the ROS environment if you haven't done so, and build the package by running the following lines:

cd ..
source /opt/ros/noetic/setup.bash
catkin build
source /devel/setup.bash

Now that we have setup the ROS package, it is time to setup the python virtual environment for the python dependencies. For simplicity, make sure that the location of your virtual environment is NOT in your workspace, so that catkin does not try to build your virtual environment. After choosing a location of desire, run the following commands to create and activate your virtual environment:

python -m venv PATH_TO_YOUR_VIRTUAL_ENV
source PATH_TO_YOUR_VIRTUAL_ENV/bin/activate

Now we install the dependencies for python in this virtual environment so it does not interfere with any other projects:

pip install numpy
pip install opencv-python
pip install ultralytics
pip install roslibpy
pip install scipy

If you want to rebuild the ROS package, you should first deactivate the python virtual environment, than clean the build and re-build the package. Run the following commands in the workspace directory to achieve this:

deactivate
catkin clean
catkin build
source devel/setup.bash

Conda/Mamba Instructions

Install this ROS Package for Ubuntu

Install a conda build of ROS and the dependencies for this product detector package by using this command:

mkdir YOUR_WORKSPACE_NAME/src --parents
cd YOUR_WORKSPACE_NAME/src
git clone [e-mail]:stijnla/Ahold_product_detector.git
cd Ahold_product_detector
conda env create -f environment.yml 

After installation, initialize a workspace in order to build and run the ROS package. Do this by running the following commands:

cd YOUR_WORKSPACE_NAME
conda activate ros_env
catkin init

We want to have our ROS package inside the 'src' directory, so now that we are in here we can clone the repository:

cd src
git clone [e-mail]:stijnla/Ahold_product_detector.git

If you want to rebuild the ROS package you should clean the build directory and re-build the package. Run the following commands in the workspace directory to achieve this:

conda activate ros_env
catkin clean
catkin build

Run the code

In order to run the code, make sure that the ROS package is built and sourced. Then run the following command with a realsense camera attached to your system:

roslaunch ahold_product_detection detect.launch

For just debugging without a robot baseframe, use

roslaunch ahold_product_detection detect_without_robot.launch

And run the following if only the detection nodes are necessary if you are using a robot for example:

roslaunch ahold_product_detection robot.launch