DOG BREED CLASSIFICATION USING DEEP LEARNING APPROACH
Loading...
Files
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Malaysia Sarawak (UNIMAS)
Abstract
Description
Identifying dog breeds from images is crucial for understanding their specific traits, such as behaviour, health, and care needs. However, distinguishing similar or mixed breeds remains challenging. To tackle this issue, a web-based application utilising deep learning approach is proposed for dog breed classification. The proposed system employs transfer learning, adapting a pre-trained EfficientNetB3 model (initially trained on ImageNet) to the Kaggle Dog Breed Dataset. This approach ensures high accuracy and efficiency in breed identification. The web application utilises computer vision technology to process user-uploaded dog images and predict the top three similar breeds, along with their corresponding similarity percentages. It offers a user-friendly interface, built with Django for the backend and HTML, CSS, and JavaScript for the frontend, ensuring accessibility and responsiveness. TensorFlow's pre-trained models are employed to enhance the system's deep learning capabilities, enabling precise and efficient classification. The system is designed to benefit dog owners, breeders, and veterinarians by providing accurate breed predictions and valuable insights. This project showcases the potential of deep learning in solving real-world classification problems. Future improvements may include features such as mixed-breed analysis, multilingual support, and integration with pet adoption databases, broadening the system's usability. By offering an innovative approach to dog breed classification, this application aims to make a meaningful contribution to AI-driven image recognition and its practical applications.