Contact person: AI Breed Recognizer for Indian Livestock

Phone:Show

Email:Show

Location: Strumica, Macedonia

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"Project Title

AI-Powered Mobile & Web App for Indian Cattle and Buffalo Breed Recognition

Project Overview

We are looking for an experienced developer or team to build a mobile and web application that uses artificial intelligence and image recognition to identify the breed of Indian cattle and buffaloes from photographs. The app will be used by Field Level Workers (FLWs) during animal registration in the Bharat Pashudhan App (BPA) to reduce breed-entry errors and improve data quality.

Objectives

Allow FLWs to capture or upload animal photos (side, front, full-body).

Use a pre-trained machine learning model to recognize and classify the breed (initially 12–20 common Indian breeds and crosses).

Handle varied backgrounds, lighting, and animal poses.

Display the top three breed predictions with confidence scores and sample images.

Provide an “Unknown / Crossbred” suggestion if confidence is low.

Enable FLWs to confirm, retake, or manually select the breed; send corrections back to the server for continuous model improvement.

Integrate seamlessly with the BPA platform via REST API to submit predicted breed and user confirmation.

Work offline if possible (lightweight on-device model); otherwise upload images to a secure server for inference.

Provide an admin dashboard to view usage metrics, model performance, and download correction logs.

Ensure secure authentication and HTTPS data transfer.

Technical Requirements

Back end: Python (FastAPI/Flask) with a machine learning model served via REST endpoints.

Model: EfficientNet/ResNet/Vision Transformer fine-tuned on an Indian cattle & buffalo breed dataset.

Mobile app: Android (Kotlin/React Native/Flutter) communicating with backend and BPA’s API.

Export the model as TFLite/ONNX for on-device inference if needed.

Scalable architecture so new breeds can be added easily.

Deliverables

Fully functional Android app (and optional web interface) implementing the above features.

REST API endpoints for breed prediction and feedback.

Simple admin panel for monitoring.

Documentation on how to retrain/update the model with new images.

Skills Required

Mobile app development (Android, optionally Flutter/React Native)

Backend/API development in Python

Machine learning (image classification, model deployment)

UI/UX design for non-technical users

Experience with secure, scalable cloud deployment

Goal

Reduce breed-entry errors in BPA by providing real-time, AI-based breed recognition and an easy-to-use workflow for field workers." (client-provided description)