README.md

metadata

title: Soil Classification Backend
emoji: 🌱
colorFrom: green
colorTo: yellow
sdk: docker
app_port: 7860
pinned: false
license: mit
tags:
  - soil-classification
  - cnn
  - fastapi
  - agriculture
  - tensorflow
  - crop-recommendation
short_description: Classify soil types from images and get crop recommendations

🌱 Soil Classification Backend

AI-powered soil image classifier built with FastAPI + TensorFlow. Upload a photo of soil and instantly receive:

• 🏷️ Soil type — one of Black · Clayey · Loamy · Red · Sandy
• 🌾 Crop recommendations — sourced from 8 000+ soil-to-crop mappings
• 🌐 Multilingual explanations — English, Hindi, and Odia

---

🔌 API

POST /api/classify

Request

{
  "imageBase64": "data:image/jpeg;base64,...",
  "language": "English"
}

Response

{
  "soilType": "Sandy",
  "explanation": "This soil appears to be Sandy soil...",
  "recommendedCrops": ["Barley", "Maize", "Millets"]
}

GET /

Health check — returns model status and supported soil classes.

---

🧠 Model

Architecture	CNN (Convolutional Neural Network)
Framework	TensorFlow / Keras 2.16
Input	224 × 224 RGB image
Classes	Black, Clayey, Loamy, Red, Sandy

---

🛠️ Stack

Web Framework	FastAPI
ML	TensorFlow 2.16 (CPU)
Image Processing	Pillow
Data	pandas + CSV (8 000+ crop mappings)
Server	Uvicorn
Container	Docker

---

📂 Files

├── app.py               # FastAPI app + CNN inference
├── Dockerfile            # Docker build
├── requirements.txt      # Python deps
├── model/
│   └── soil_model007.keras
└── data/
    └── data_core.csv     # Soil → Crop mapping

---

🌍 Languages Supported

Language	Code
English	"English"
Hindi	"Hindi"
Odia	"Odia"

---

Made with ❤️ for Indian agriculture · NIELIT Internship