Sčítač dopravy

Scripts for video-based automated counting and reporting of vehicle traffic. The yolov8 model is used for detecting and tracking vehicles based on a live video stream. Traffic data is aggregated and a report compatible with TP 189 (Czech standard for measuring road traffic intensity) can be generated.

Installation

python -m venv .venv
source .venv/bin/activate

pip install -r requirements.txt

Configuration

Detection script configuration is by default read from the config-live.yaml file. The file provides reasonable defaults for detecting vehicles in a live video stream. The detection region should be adjusted based on the given location.

Vehicle detection

Start the detection script by running:

python detect.py <live stream URL>

A preview of the augmented video stream will pop up.

An empty file is created in workdir every time a vehicle is detected. The file name contains a timestamp of when the detection occurred. The files are organized in nested folders based on the date.

Generating reports

Reports can be generated from information stored in workdir.

python evaluate.py

2024-02-05,08:45,09:45,113
2024-02-06,08:15,09:15,117
2024-02-07,08:45,09:45,124
2024-02-08,08:00,09:00,138
2024-02-09,08:00,09:00,126

By default, for each day the script identifies the peak hour (4 subsequent 15-minute blocks with highest traffic) within a given window (7:00-17:00 by default). Alternatively, a full report which includes the traffic count for each 15-minute block can be generated:

python evaluate.py -f full

2024-02-05,08:30,08:45,14, 
2024-02-05,08:45,09:00,35, 
2024-02-05,09:00,09:15,22, 
2024-02-05,09:15,09:30,21,92
2024-02-05,09:30,09:45,35,113
2024-02-05,09:45,10:00,20,98
2024-02-05,10:00,10:15,28,104
2024-02-05,10:15,10:30,19,102
2024-02-05,10:30,10:45,27,94
2024-02-05,10:45,11:00,30,104

The last column contains the traffic sum for the corresponding "floating hour" i.e. sum of the given 15-minute block plus three immediatelly preceeding blocks (if available).

Uploading reports to Google Sheets

The reports can be directly uploaded to Google Sheets:

Follow py_gsheets documetation to obtain a service account key.
Run the evaluate script

python evaluate.py -d 'name of the Google Sheet' -saf 'path-to-service-account-key' -s 'sheet number'