Classification of Dog Breeds Using Convolutional Neural Network Models and Support Vector Machine.

Area of Science:

• Computer Vision
• Machine Learning
• Zoology

Background:

• Accurate dog breed classification is crucial for identification and research.
• Traditional methods struggle with the diversity and similarities among dog breeds.
• Convolutional Neural Networks (CNNs) offer advanced feature learning but face challenges with breed diversity.

Purpose of the Study:

• To develop an advanced model for significantly improving dog image classification accuracy.
• To overcome the limitations of existing methods in distinguishing between diverse dog breeds.

Main Methods:

• Integration of multiple CNN models for feature extraction.
• Application of Principal Component Analysis (PCA) and Gray Wolf Optimization (GWO) for feature filtering.
• Classification using Support Vector Machine (SVM) on processed features.

Main Results:

• Achieved 95.24% accuracy for 120 dog breeds.
• Reached 99.34% accuracy for a subset of 76 selected breeds.
• Demonstrated superior performance compared to existing methods on the Stanford Dog Dataset.

Conclusions:

• The proposed integrated model significantly enhances dog breed classification accuracy.
• This method provides a robust framework for classifying a wide range of species.
• The approach offers a substantial advancement in automated image-based species identification.