🐄 Cattle Breed Classification System (CattleAI)

An end-to-end deep learning–based web application that identifies cattle breeds from images using a Convolutional Neural Network (CNN) with transfer learning, deployed via a Flask web interface.

📌 Project Overview

Cattle breed identification is a challenging fine-grained image classification problem due to high visual similarity between breeds. This project leverages deep learning and transfer learning to automatically classify cattle breeds from uploaded images and display predictions through a user-friendly web interface.

The system allows users to upload a cattle image and instantly receive: Predicted breed name Prediction confidence

🚀 Features

🧠 CNN-based cattle breed classification
🔁 Transfer learning with fine-tuning
🛑 Overfitting control using EarlyStopping and learning-rate scheduling
🌐 Web interface built using HTML, CSS, JavaScript, and Bootstrap
⚙️ Flask backend for model inference
📊 Model evaluation using confusion matrix and classification report

🧠 Cattle Breeds Classified

The model is trained to classify the following breeds: 1. Ayrshire cattle 2. Brown Swiss cattle 3. Holstein Friesian cattle 4. Jersey cattle 5.Red Dane cattle

🛠️ Tech Stack Machine Learning & Backend

Python

TensorFlow / Keras

NumPy

Pillow (PIL)

Flask

SciPy

Frontend

HTML5

CSS3

JavaScript

Bootstrap 5

Tools

VS Code

Git & GitHub

🔄 Application Workflow

This diagram illustrates the end-to-end workflow of the application, from image upload to final cattle breed prediction.

Architecture Diagram

🧪 Model Training Details

Approach: Transfer Learning

Base Model: Pretrained CNN (MobileNet / similar)

Training Strategy:

Freeze pretrained layers (feature extraction)

Train custom classification head

Fine-tune top layers with low learning rate

Overfitting Prevention Techniques

EarlyStopping

ReduceLROnPlateau

Dropout layers 

Data augmentation

📊 Model Evaluation

The model was evaluated using:

Accuracy

Confusion Matrix

Precision, Recall, and F1-score

Web Application Workflow

User uploads a cattle image

Image is sent to Flask backend

Image is processed using Pillow

CNN model predicts breed

Result is returned and displayed on the webpage

▶️ How to Run the Project Locally

1. Clone the Repository

    git clone https://github.com/your-username/Classification_proj.git

    cd Classification_proj

2. Create and Activate Virtual Environment
   
    python -m venv .venv
   
    source .venv/Scripts/activate   # Windows

3. Install Dependencies
    
    pip install -r requirements.txt

🔮 Future Enhancements

Collect larger and more diverse datasets

Use stronger architectures (EfficientNet, ResNet)

Improve accuracy using attention mechanisms

Deploy application online (Render / Railway)

Convert model to TensorFlow Lite

🎯 Learning Outcomes

Built an end-to-end ML pipeline

Gained hands-on experience with CNNs

Learned transfer learning and fine-tuning

Understood overfitting control techniques

Integrated ML models into web applications

🖥️ Web Application Output

Home Page

Prediction Result

The application displays the predicted cattle breed along with confidence score.