• Python 3.7;
• Official NVIDIA driver;
• CUDA packages.

• Open a terminal window and run the following commands to remove any NVIDIA/CUDA packages the system may already and that may cause a conflicts with newly installed ones:

sudo rm /etc/apt/sources.list.d/cuda*
sudo apt remove --autoremove nvidia-cuda-toolkit
sudo apt remove --autoremove nvidia-*

• Purge any remaining NVIDIA configuration files and associated dependencies that may have been installed with:

sudo apt-get purge nvidia*
sudo apt-get autoremove
sudo apt-get autoclean

• Remove any existing CUDA folders the system may have in /user/local/

sudo rm -rf /usr/local/cuda*

• Set CUDA PPA:

sudo apt update
sudo add-apt-repository ppa:graphics-drivers
sudo apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/<UBUNTU_VERSION>/x86_64/7fa2af80.pub
sudo bash -c 'echo "deb http://developer.download.nvidia.com/compute/cuda/repos/<UBUNTU_VERSION>/x86_64 /" > /etc/apt/sources.list.d/cuda.list'
sudo bash -c 'echo "deb http://developer.download.nvidia.com/compute/machine-learning/repos/<UBUNTU_VERSION>/x86_64 /" > /etc/apt/sources.list.d/cuda_learn.list'

Replace <UBUNTU_VERSION> with the proper version of the Ubuntu, ex. ubuntu1804.

And finally install the packages:

sudo apt update
sudo apt install cuda-<MAJOR_VERSION_NUMBER>-<SECONDARY_VERSION_NUMBER-number>
sudo apt install libcudnn7

• OR follow the steps from the official documentation page here on installation of the CUDA packages of the desired version

Also, Check out the CUDA installation guide for Linux here.

After installing, we need to add CUDA to our PATH, so that the shell knows where to find CUDA. To edit our path, open up the ‘.profile’ file using vim or nano.

sudo vim ~/.profile

And add these lines to the end of the file:

# set PATH for cuda installation
if [ -d "/usr/local/cuda-<MAJOR_VERSION_NUMBER>.<SECONDARY_VERSION_NUMBER>/bin/" ]; then
    export PATH=/usr/local/cuda-<MAJOR_VERSION_NUMBER>.<SECONDARY_VERSION_NUMBER>/bin${PATH:+:${PATH}}
    export LD_LIBRARY_PATH=/usr/local/cuda-<MAJOR_VERSION_NUMBER>.<SECONDARY_VERSION_NUMBER>/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
fi

Replace <MAJOR_VERSION_NUMBER> and <SECONDARY_VERSION_NUMBER> with appropriate CUDA version numbers.

Reboot the system to make the changes take an effect.

sudo reboot

Check if the NVIDIA driver and the CUDA packages have been installed successfully.

• Check NVIDIA driver:

nvidia-smi

• Check CUDA:

nvcc --version

• Check CuDNN:

/sbin/ldconfig -N -v $(sed ‘s/:/ /’ <<< $LD_LIBRARY_PATH) 2>/dev/null | grep libcudnn

NOTE: CuDNN is not required for the successful run of the Vehicle Counting Software.

Reference: https://medium.com/@stephengregory_69986/installing-cuda-10-1-on-ubuntu-20-04-e562a5e724a0

• Clone this repo [email protected]:Artem723/ivy.git.
• Create a virtual environment:

python -m venv <PATH>

Use the virtual environment:

source <PATH>/bin/activate

<PATH> - is the place where the virtual environment and packages will be created and installed.

• Run pip install -r requirements.txt to install dependencies.

The configuration file for the software is .env. (If it is not initially visible, then enable the possibility to see the hidden files in the file explorer) This file contains the environment variables for the software, such as the input folder, the position of the examining lines etc.

In the .env, set the path to the folder with the Miovision video files:

INPUT="<PATH_TO_VIDEO_OR_FOLDER>"

The extension of video files should be *.mp4

The examining lines (L_SPEED and R_SPEED on the image) serve for the speed estimation and counting of the vehicles. The vehicle is counted only if it passes two lines.

EXAMINING_LINES = [{'label': 'L_SPEED', 'start': (<L_X>, <L_Y>), 'length': <L_LENGTH>}, {'label': 'R_SPEED', 'start': (<R_X>, <R_Y>), 'length': <R_LENGTH>}, {'label': 'C', 'start': (300, 0), 'length': 400}]

The <L_X>, <L_Y> and <R_X>, <R_Y> position of the beginning point of the examining line on the video frame. On the picture they have coordinates (20; 20) and (450; 20).

The <L_LENGTH>\ and <R_LENGTH>\ is the length in pixels of the examining lines. On the picture, the length is 300px.

In order to better understand the position of the lines, press "P" (Pause) on keyboard to suspend the execution of the software, then left-click on the interested point of the video frame. In the terminal, you should see the relative position where you clicked:

[2024-07-18 18:00:13,935] INFO    : Pixel position captured. {'label': 'PIXEL_POSITION', 'position': (460, 310)}

Here, (460, 310) the position where the user clicked over the video frame.

To resume the processing, press "P" again.

Assign the real distance between the examining lines in meters to the "DISTANCE_BETWEEN_SPEED_LABELS" parameter:

DISTANCE_BETWEEN_SPEED_LABELS = 6.5 # in METERS

This parameters serves primarily for the speed estimation.

• Run python -m main.

Download vcs_demo_data.zip and unzip its contents in the data directory. It contains detection models and a sample video of a traffic scene.

python -m pytest

By default, the VCS runs in "debug mode" which provides you a window to monitor the vehicle counting process. You can press the s key when the program is running to capture a screenshot and use q to quit.

The vehicle counting system is made up of three main components: a detector, tracker and counter. The detector identifies vehicles in a given frame of video and returns a list of bounding boxes around the vehicles to the tracker. The tracker uses the bounding boxes to track the vehicles in subsequent frames. The detector is also used to update the trackers periodically to ensure that they are still tracking the vehicles correctly. The counter counts vehicles when they leave the frame or makes use of a counting line drawn across a road.

PS: You can find out about how the vehicle counting system was built by checking out this article on my blog: https://alphacoder.xyz/vehicle-counting/.

https://ai.stanford.edu/~jkrause/cars/car_dataset.html https://sites.google.com/view/visionlearning/databases/image-database-for-vehicle-recognition https://public.roboflow.com/object-detection/vehicles-openimages/1/download https://lionbridge.ai/datasets/250000-cars-top-10-free-vehicle-image-and-video-datasets-for-machine-learning/