This programm was developed 2024 as a prototype for my bachelor's thesis Prototyp eines Food Scanners zur automatischen Erkennung von Speisen und Ermittlung ihrer Nährwerte. It consists of some Jupyter Notebooks for preparing an image recognition model and a webapp. The webapp was developed to identify images of food in real-time and display it's Nutri-Score. The Nutri-Score is calculated in this repository. The evaluate.ipynb notebook was used to generate some graphics regarding the evaluation and to measure the speed of the webserver.
With a python 3.10 environment installed and activated execute:
pip install -r app/requirements.txt
Then initialize the DB with:
python app/foodscanner/dbms/dbms.py
Now fill the database with your food data e.g.:
Start the webserver with :
python app/foodscanner/web/server.py --model_path=< path to your model's weights file>
The server should now run under the local url. To test different layout variations try one of:
Download the datasets with:
Train a model with:
Start the Webapp as described in Run.