This project is designed to be a simple solution for a smart parking system.
The first component is a script that allows easy (but manual) labelling of parking spots with a selection-tool.
The second component automatically picks up those labelled parking spots from a stored location, and detects whether a vehicle is parked in those spaces or not.
This allows a simple and interactive way of determining free spots in a parking area.
First setup the environment, and install requirements. Ensure PyTorch is installed with CUDA/GPU enabled for faster processing.
Start off by running the select-regions.py script, and providing an image input (a screenshot/still frame) from the video feed of a parking lot. Then, follow the instructions and draw in the bounding regions for the parking spots.
Next, run the smart-parking.py script after modifying the parameters as per specific needs, and see the program automatically detecting which spaces are free/occupied. Works with live feed or a video file input.
This project was developed with the following setup:
- Python 3.11+
- PyTorch 2.4.1 with CUDA 12.4