Dog Vision: AI-Powered Dog Breed Classification 🐶

Welcome to the Dog Vision project! This project leverages deep learning techniques to classify dog breeds from images using a pre-trained convolutional neural network (CNN). Built with Python and TensorFlow, Dog Vision demonstrates how AI/ML can be applied to real-world image classification problems.

Table of Contents 📑

• Overview
• Model
• Results

Overview

The Dog Vision project aims to classify dog breeds from images by applying deep learning. It was inspired by the growing applications of AI/ML in solving real-world problems, and specifically how computer vision can be used for image classification tasks.

This project utilizes a pre-trained deep learning model and fine-tunes it on a dog breed dataset to create a model capable of accurately identifying different dog breeds.

Key features:

• Uses a pre-trained Convolutional Neural Network (CNN) for feature extraction and classification.
• Trains on a dog breed dataset and evaluates model performance.
• Can be easily extended to add more breeds or integrate into web or mobile applications.

Try It in Google Colab 🚀

You can explore the project and test the model directly in Google Colab here.

Model

The model is based on a pre-trained convolutional neural network (CNN). The model was fine-tuned on a dog breed dataset using TensorFlow and Keras. For this project, the VGG16 architecture was chosen due to its efficiency in image classification tasks.

Training:

• Dataset: A large dataset of images of various dog breeds.
• Model Architecture: VGG16 with fine-tuning.
• Optimizer: Adam Optimizer.
• Loss Function: Sparse Categorical Crossentropy.

Accuracy:

• The model achieved an accuracy of 95% on the test set.

Results 📊

• Accuracy: 95%
• Top 5 Predicted Breeds:
  1. Golden Retriever
  2. German Shepherd
  3. Bulldog
  4. Beagle
  5. Poodle

You can try your own images to see how well the model performs.