conda create -f dev-environment.yml
Download the dataset from here in the Yolov8 format and put the content into the License-Plates-5
directory.
You can train the classifier using the provided jupyter notebook training.ipynb
.
In order to finetune yolo, you need to provide the yolo model as yolov5s.pt
.
The repository contains the fine-tuned model as finetuned-model.pt
.
The repository contains a webapp for demonstration puroposes.
The backend files can be found in the licenseplates
module. It is built upon fastapi.
The frontend is built upon react and can be found in licensenplates-fe
directory.
You can start frontend and backend using docker compose
:
docker compose up
The webapp is then available at http://localhost:3000
.
In the following, you can see how the webapp looks like:
You can also use the predict.py
script to manual predict a license plate:
python predict.py