Weed Detection Backend

An intelligent Weed Detection Backend built using FastAPI, TensorFlow, and PyTorch, designed to classify and segment images for automated weed identification.
This repository serves as the backend API powering the weed detection system, capable of handling image uploads, preprocessing, and model inference across multiple deep learning architectures.

---

Features

• Supports multiple model architectures: U-Net, ResNet50, EfficientNetB0, and custom TensorFlow models
• Fast inference with FastAPI and Uvicorn
• Handles image uploads and validation (JPEG/PNG)
• Configurable model registry and environment
• Built-in Pydantic schemas for request/response validation

---

Tech Stack

Category	Technologies
Backend Framework	FastAPI
Deep Learning	TensorFlow, PyTorch, TorchVision
Data Handling	Pillow, NumPy, Pydantic
Server	Uvicorn
Environment	Python 3.9+

---

Project Structure

Weed_Detection_Backend/
│
├── app/
│   ├── main.py                 # FastAPI entry point
│   ├── config.py               # Configuration (paths, constants, etc.)
│   ├── schemas.py              # Pydantic models for API validation
│   ├── utils/
│   │   ├── registry.py         # Handles model loading and registration
│   │   └── __init__.py
│   └── models/
│       ├── resnet.py           # ResNet50 model handler
│       ├── efficientnet.py     # EfficientNetB0 model handler
│       ├── unet.py             # U-Net segmentation model handler
│       └── model4.py           # Custom TensorFlow model
│
├── .github/
│   └── workflows/
│       └── main_weeddetection.yml   # Azure CI/CD pipeline
│
├── requirements.txt            # Project dependencies
└── README.md                   # Project documentation (you are here)

---

Installation & Setup

Clone the repository

git clone https://github.com/Pyro-Warrior-1884/Weed_Detection_Backend.git
cd Weed_Detection_Backend

Create a virtual environment

python -m venv venv
source venv/bin/activate      # For Linux/macOS
venv\Scripts\activate         # For Windows

Install dependencies

pip install -r requirements.txt

Run the FastAPI server

uvicorn app.main:app --reload

Access the API

Visit:

http://127.0.0.1:8000

---

Model Overview

Model	Framework	Input Size	Purpose
U-Net	TensorFlow	128×128	Image segmentation & weed area detection
ResNet50	TensorFlow	224×224	Image classification
EfficientNetB0	TensorFlow	224×224	Lightweight classification
Model4	TensorFlow	224×224	Custom model template

---

API Schemas

ModelInfo

{
  "id": "string",
  "name": "ResNet50",
  "framework": "tensorflow",
  "input_size": 224
}

PredictionResponse

{
  "model": "unet",
  "prediction": "weed_detected",
  "confidence": 0.93
}

---

Testing

You can test predictions using cURL or Postman:

curl -X POST "http://127.0.0.1:8000/predict" -F "file=@sample.jpg"

Expected response:

{
  "model": "unet",
  "prediction": "weed_detected",
  "confidence": 0.94
}

---

Configuration

Located in app/config.py

BASE_DIR = "/path/to/project/base/dir"
MODEL_DIR = BASE_DIR + "/models"
MAX_UPLOAD_SIZE = 1024 * 1024  # 1 MB
ACCEPTED_CONTENT_TYPES = ["image/jpeg", "image/png"]
ALLOWED_ORIGINS = ["*"]

You can modify these constants to adjust:

• Upload limits
• Accepted MIME types
• Allowed CORS origins

---

Notes & Considerations

• Ensure model files (e.g., unet.h5, resnet50.h5) are present in the models/ directory.
• Use consistent input image sizes (224x224 or 128x128).
• Fine-tune threshold values in segmentation for better weed detection accuracy.

---

Contributing

Contributions are welcome!
To contribute:

1. Fork the repository
2. Create a new branch (feature/my-feature)
3. Commit your changes
4. Submit a Pull Request

---

Author

Albert Augustine
GitHub: @Pyro-Warrior-1884
Reach out for collaborations or ML-based backend integrations.

---

Don’t forget to star this repo if you find it useful!