Backend AI Guide

FastAPI backend to classify images using a neural network (Keras/TensorFlow). The service receives an image, processes it with a pre-trained model, and returns the identified category together with a related audio file, which is served as a static asset.

Features

• FastAPI: quick API with automatic docs at /docs.
• TensorFlow/Keras: loads model (modelo.h5) and weights (pesos.h5).
• Image preprocessing: resize and array conversion via Pillow/Keras utils.
• Static files: audios served from /audios.
• Configurable CORS: allowed origins via environment variables.

Project structure

backend-AI-guide/
├─ main.py
├─ modelo/
│  ├─ modelo.h5
│  └─ pesos.h5
├─ media/
│  ├─ audios/
│  │  ├─ cacao.mp3
│  │  ├─ metate.mp3
│  │  ├─ molinillo.mp3
│  │  ├─ mortero.mp3
│  │  └─ silla.mp3
│  └─ uploads/
│     └─ 205f08ba-e978-4189-9d9c-21c125e1e7cb.jpg
├─ requirements.txt
└─ README.md

Requirements

• Python 3.10+
• pip / venv

Installation

# 1) Create and activate a virtual environment (recommended)
python -m venv venv
# Windows PowerShell
venv\\Scripts\\Activate.ps1

# 2) Install dependencies
pip install -r requirements.txt

Environment variables

Create a .env file in the project root to configure CORS:

ORIGIN_FRONTEND=http://localhost:5173
HOST_ORIGIN_FRONTEND=http://0.0.0.0:5173

Adjust the values to match your frontend origins.

Run

# Run in development
fastapi dev main.py --host 0.0.0.0 --port 8000

• API base: http://127.0.0.1:8000
• Swagger docs: http://127.0.0.1:8000/docs
• Static files (audios): http://127.0.0.1:8000/audios/<filename>

Endpoints

• GET / – Basic service status.
• POST /uploadfile – Accepts an image file (multipart/form-data) and returns the prediction.

Example request (cURL)

curl -X POST "http://127.0.0.1:8000/uploadfile" \
  -H "accept: application/json" \
  -H "Content-Type: multipart/form-data" \
  -F "file=@path/to/your_image.jpg"

Example response

{
  "data": {
    "nombre": "Cacao",
    "audio": "/audios/cacao.mp3"
  }
}

Model

Place your model and weights at modelo/modelo.h5 and modelo/pesos.h5. The backend will load them when processing the image. Make sure they match the preprocessing used during training (size 100x100 according to main.py).

Media and uploads

• Uploaded images are saved to media/uploads/ with a unique filename.
• Audios are publicly served from /audios (mapped to media/audios/).

Error handling

• The service returns clear messages when image loading/reading, model loading, or prediction fails.
• Check the server console logs for more details if errors occur.

Development

• Interactive documentation: visit /docs when the server is running.
• Adjust CORS origins in .env to allow requests from your frontend.

License

MIT License

Copyright (c) 2025 Jeremy Alexander Ramírez Galeotti