Dog Breed Classifier

Project Overview

The Dog Breed Classifier is a web application that uses deep learning to identify dog breeds from user-uploaded images. Built with Flask and powered by a MobileNetV2 model fine-tuned on the Stanford Dogs dataset, the application delivers real-time predictions through a user-friendly drag-and-drop interface.

Objective

The aim of this project is to demonstrate the practical use of convolutional neural networks (CNNs) for image classification and deliver the results through a fully integrated web-based interface. It combines deep learning inference with web deployment techniques in a streamlined, production-ready application.

Tools and Technologies

• Programming Language: Python
  
  

• Frameworks and Libraries: Flask, Keras, TensorFlow, OpenCV
  
  

• Model Architecture: MobileNetV2 (fine-tuned)
  
  

• Frontend: HTML, CSS, JavaScript (AJAX), Bootstrap
  
  

• Deployment: Render
  
  

• Containerization: Docker (optional)
  
  

Key Features

• Upload an image to receive an instant prediction of the dog breed
  
  

• Responsive UI with drag-and-drop image upload and preview
  
  

• Backend serves both frontend and prediction API via Flask
  
  

• Uses a fine-tuned MobileNetV2 model for efficient and accurate inference
  
  

• Includes Docker support for easy container-based deployment
  
  

• Cloud deployment ready for platforms like Render, Railway, and Heroku
  
  

Application Workflow

1. User uploads a dog image via the drag-and-drop UI
   
   

2. The image is preprocessed and passed through the MobileNetV2 model
   
   

3. The predicted breed is returned and displayed on the interface
   
   

4. Uploaded files are temporarily stored and cleaned after use