AgriNN

AI-Powered Animal Type Classification for Precision Dairy Farming

AgriNN is an AI-powered livestock classification system that uses computer vision to analyze images of cattle and buffaloes. It extracts body structure features such as body length, height, and key proportions, then generates standardized classification scores to support Animal Type Classification (ATC).

Repository: https://github.com/adityai0/AgriNN

Problem Statement

Manual livestock type assessment is time-consuming, subjective, and prone to human bias, which can reduce consistency and accuracy in dairy farming and breeding decisions.

Solution Overview

AgriNN automates ATC using an end-to-end AI pipeline. Users upload animal images, the system detects whether the subject is cattle or buffalo, extracts structural body features, computes standardized scores, and presents results with visual overlays. Historical records are stored for review and long-term analysis.

Features

• Upload animal images from a web interface
• Detect cattle or buffalo using AI-based object detection
• Extract body structure features (length, height, proportions)
• Generate standardized classification scores
• Display results with visual overlays for interpretability
• Store and retrieve historical analysis records

Tech Stack

• Frontend: Next.js, Tailwind CSS
• Backend: FastAPI (Python)
• AI/ML: PyTorch, YOLOv8, OpenCV
• Database: PostgreSQL

System Architecture

[User]
	|
	v
[Next.js + Tailwind Frontend]
	|
	| HTTP/REST
	v
[FastAPI Backend]
	|            \
	|             \--> [PostgreSQL: records, metadata, scores]
	v
[Inference Pipeline: YOLOv8 + OpenCV + PyTorch]
	|
	v
[Feature Extraction + ATC Scoring Engine]
	|
	v
[API Response: class, measurements, score, overlays]

Installation

Prerequisites

• Node.js 18+
• Python 3.10+
• PostgreSQL (local or cloud)
• Git

1. Clone the Repository

git clone https://github.com/adityai0/AgriNN.git
cd AgriNN

2. Backend Setup (FastAPI)

# from project root
cd backend
python -m venv .venv

# Windows
.venv\Scripts\activate

# Linux/macOS
source .venv/bin/activate

pip install -r requirements.txt

Create a .env file in backend/ with values similar to:

DATABASE_URL=postgresql://postgres:postgres@localhost:5432/agrinn
MODEL_PATH=./models/yolov8.pt

Start backend:

uvicorn app.main:app --reload --host 0.0.0.0 --port 8000

3. Frontend Setup (Next.js)

# from project root
cd frontend
npm install

Create a .env.local file in frontend/:

NEXT_PUBLIC_API_BASE_URL=http://localhost:8000

Start frontend:

npm run dev

Frontend runs at http://localhost:3000.

Usage

1. Open the frontend in your browser.
2. Upload an image of cattle or buffalo.
3. Submit for analysis.
4. Review detected animal type, extracted measurements, and classification score.
5. Inspect visual overlays and save/revisit historical records.

Future Improvements

• Multi-animal image handling and tracking
• Mobile-first capture flow for on-farm use
• Explainable AI dashboard with confidence decomposition
• Breed-specific calibration of scoring rules
• Role-based access and audit logging
• Model monitoring and auto-retraining workflows

Contributing

Contributions are welcome. Please read CONTRIBUTING.md for setup, coding standards, and pull request guidelines.

License

This project is licensed under the MIT License. See LICENSE for details.