razyoboy / food_for_thoughts

Tensorflow-based ML model for food classification and detection

Food for Thoughts

A Tensorflow-based ML model for food classification and detection, based on the Kaggle dataset food-101 and thfood-50 - utilizing the InceptionV3 Convolutional Neural Network (CNN) model.

Usage (via API)

This can be done locally, or via a Cloud serverless provider of your choice. In our case, we used Google Cloud Platform as the main Cloud provider.

Deployment via Google Cloud Run (CLI-method)

1. Install gcloud CLI and follow the instructions as shown here
2. Clone this repository and navigate to its root directory

git clone https://github.com/razyoboy/food_for_thoughts/
cd food_for_thoughts

3. Deploy to your project by using the following command (more options can be found here)

gcloud run deploy

Deployment via Local PC

1. Clone this repository and navigate to the root directory

git clone https://github.com/razyoboy/food_for_thoughts/
cd food_for_thoughts

2. Ensure all requirements are met

pip install -r requirements.txt

3. Run the local server

python -u "src/main.py"

API Endpoints and Interaction

Currently, food_for_thoughts supports two endpoints:

GET Check Health

• <base_url>/api/

Returns

{
    "status": 200,
    "text": "Up and Running!"
}

POST Predict Food

• <base_url>/api/predict-food/
• form-data body with image-file as the key
• Attach image via the image-file key

Returns

{
 "status": 200,
 "prediction_results": {
     "food_type": <food_result>,
     "confidence": <percentage>
 }
}

_{DISCLAIMER: This project is part of EGBI443 Image Processing in Medicine class project, and is not intended to be a production ready implementation for food prediction via API calls}