Skip to content

API Documentation

Overview

This API provides functionality for classifying images as hotdog or not-hotdog. The service loads a pre-trained model and performs predictions based on uploaded images.

Request URL: https://api-service-748263339953.europe-west1.run.app/

Interactive URL https://api-service-748263339953.europe-west1.run.app/docs

Endpoints

1. Root Endpoint

Description: Returns a basic message to confirm the API is running.

  • Method: GET
  • URL: /

Example Request: 

curl -X 'GET' \
  'https://api-service-748263339953.europe-west1.run.app/' \
  -H 'accept: application/json'
Response: 
{
    "message": "Image Classification API is running"
}

2. Prediction Endpoint

Description: Classifies an uploaded image as either hotdog or not-hotdog.

  • Method: POST
  • URL: /predict
  • Headers:
  • Content-Type: multipart/form-data
  • Request Body:
  • file (required): The image file to classify. Must be .jpg, .jpeg, or .png.

Example Request: 

curl -X 'POST' \
  'https://api-service-748263339953.europe-west1.run.app/predict' \
  -H 'accept: application/json' \
  -H 'Content-Type: multipart/form-data' \
  -F 'data=@your_image.jpg;type=image/jpeg'

Response: 

{
    "category": "hotdog",
    "probabilities": 0.95
}
- category: The classification result (hotdog or not-hotdog). - probabilities: The confidence score for the prediction.

Error Responses: - Invalid file format: 

{
    "detail": "Invalid file format. Only JPG and PNG images are supported."
}
- Model not loaded: 
{
    "detail": "Model is not loaded. Please try again later."
}
- Internal server error: 
{
    "detail": "Error processing image: <error_message>"
}

3. Health Check Endpoint

Description: Checks the health status of the API and confirms if the model is loaded.

  • Method: GET
  • URL: /health

Example Request 

curl -X 'GET' \
  'https://api-service-748263339953.europe-west1.run.app/health' \
  -H 'accept: application/json'

Response: 

{
    "status": "healthy",
    "model_loaded": true
}

Error Response: Model not loaded: 

{
    "detail": "Model not loaded"
}