To run the file with talker/listeners

run in termnial:

roslaunch hard_hanze cv.launch

How to use/compile/build the computer vision module (Main module)

Library dependencies

• cmake >= 3.4
• C/C++14 compiler

Submodules

This project uses one submodule and we need to initialize and update it

git submodule update --init --recursive

Prerequisites

If OpenVINO is installed then no further action is required.

Otherwise install one of the followring, depending on the machine

• Ubuntu 20.04/22.04 - sudo apt install libopencv-dev
• MacOS - brew install opencv
• Or install from package / sources (Windows and other OSes) https://docs.opencv.org/master/d0/d3d/tutorial_general_install.html

Build DepthAI-Core library

cd depthai-core

git submodule update --init --recursive

Building the library

cmake -S. -Bbuild
cmake --build build

More documentation can be found at: https://github.com/luxonis/depthai-core

Then we head back to the main directory

cd ..

Building

mkdir -p build && cd build
cmake ..
cmake --build . --parallel

Runing

./detectionModule

How to use the computer vision module (Demo module)

!!! This module is just for demonstation purpose !!!

1. Installing the dependencies

Refer to the DepthAI docs: https://docs.luxonis.com/en/latest/pages/tutorials/first_steps/#first-steps-with-depthai

2. Running the module by itself

$ cd Driverless/computer_vision/YOLOv5/scripts
$ python3 cvReworked.py

3. Understanding the folder configuration

We currently work on the files in the folder "Yolov5."

$ cd Driverless/computer_vision/YOLOv5

Here we have the following structure:

|YOLOv5
|-details
|-pts
|-src
|-shaves

"details" folder

We store the JSON files of the models we test in the details folder. You can find some examples of "detail" files at the moment.

"pts" folder

In the pts folder, we store the .pt files. These files are the raw file that results from a training session for a model. To train a model, we use the following jupiter notebook: https://github.com/luxonis/depthai-ml-training/blob/master/colab-notebooks/YoloV5_training.ipynb.

For training a model on a local machine with an Nvidia GPU, refer to this doc: TBA.

After the files are trained, and it results in a .pt file, we use this online transformer: http://tools.luxonis.com/. As the input shape, put the size of the model (The default size we use is: 416)

The files resulting from the conversion are in a .zip file that will be downloaded from the site.

From all the files inside the .zip is the .json, which always goes to the "details" folder, and the .blob file, which goes to the "shaves" folder.

"src" folder

Here we have all the code that we use for the model.

The only file currently in use is "cvReworked.py." All other files are for reference. To run the code, use the following:

$ python3 cvReworked.py

"shaves" folder

Here are saved all the .blob results from the conversion of the .pt file.