Dog Vision 🐶

Dog Vision is a Deep Learning-based Dog Breed Classification project that identifies dog breeds from images using Transfer Learning and TensorFlow. The model is trained on the Kaggle Dog Breed Identification dataset containing over 10,000 images across 120 dog breeds. :contentReference[oaicite:0]{index=0}

📌 Project Overview

Identifying dog breeds from images is a Multi-Class Image Classification problem. This project leverages a pre-trained Convolutional Neural Network (CNN) through Transfer Learning to accurately classify images into one of 120 dog breeds. Transfer Learning allows us to reuse knowledge from models trained on large image datasets, improving accuracy while reducing training time. :contentReference[oaicite:1]{index=1}

🚀 Features

• Classifies 120 different dog breeds
• Built using TensorFlow 2.x and TensorFlow Hub
• Uses Transfer Learning for improved performance
• Supports image preprocessing and prediction pipelines
• Model saving and reloading functionality
• Prediction probability outputs for each breed
• End-to-end Machine Learning workflow

🛠️ Tech Stack

• Python
• TensorFlow 2.x
• TensorFlow Hub
• NumPy
• Pandas
• Matplotlib
• Scikit-Learn
• Jupyter Notebook

📂 Dataset

The project uses the Kaggle Dog Breed Identification Dataset:

• 10,000+ labeled dog images
• 120 unique dog breeds
• Multi-class classification problem

Dataset Link: https://www.kaggle.com/c/dog-breed-identification

🧠 Model Architecture

The project utilizes a pre-trained deep learning model through Transfer Learning.

Workflow

1. Data Collection
2. Data Preprocessing
3. Label Encoding
4. Transfer Learning Model Creation
5. Model Training
6. Model Evaluation
7. Prediction Generation
8. Model Saving and Loading

📊 Results

The trained model can:

• Predict dog breeds from unseen images
• Generate confidence scores for predictions
• Handle multi-class image classification effectively

📁 Project Structure