Canine Skin Disease Classification Using Convolutional Neural Networks (CNN)

Aim:

        To develop a custom Convolutional Neural Network (CNN) model for accurately classifying seven common canine skin diseases, thereby improving diagnostic precision and supporting veterinary care.

Abstract:

          Skin diseases in dogs cause discomfort and can impact human health through transmission. Existing models like MobileNetV2 and VGG-16 classify diseases into broad categories (bacterial, fungal, hypersensitivity) but struggle to distinguish specific diseases. The proposed system introduces a custom CNN model to identify seven canine skin diseases, including Pyoderma, Mange, and Flea Allergy Dermatitis. By incorporating data augmentation, class balancing, and model fine-tuning, the system aims to achieve improved accuracy and surpass the performance of existing models. This approach will enhance disease detection and improve veterinary care and pet health management.

Introduction:

     Canine skin diseases such as bacterial infections, fungal infections, and allergies are common. Current models classify diseases broadly but fail to pinpoint specific diseases. Accurate identification is crucial for timely treatment and effective veterinary care. This research proposes a custom CNN model to classify seven canine skin diseases with higher precision than existing models, enabling better diagnosis and treatment planning.

 Existing System:

      Current classification systems use models like MobileNetV2, VGG-16, and ResNet to categorize diseases into broad classes (bacterial, fungal, hypersensitivity). However, these models cannot distinguish specific diseases within these categories. While these models provide a baseline for classification, they lack the granularity needed for precise diagnosis.

Disadvantages of Existing System:

• Limited to broad disease categories.
• Inability to distinguish specific diseases.
• Relatively lower accuracy.

 Proposed System:

       The proposed system introduces a custom CNN model to classify seven specific canine skin diseases with improved accuracy. The system employs data augmentation, class balancing, and model fine-tuning techniques with MobileNetV2, InceptionV3 to enhance performance and provide more granular classification.

 Diseases Classified:

1. Bacterial Dermatosis (Pyoderma)
2. Flea Allergy Dermatitis
3. Fungal Infection
4. Hot Spots (Acute Moist Dermatitis)
5. Hypersensitivity Allergic Dermatosis
6. Mange
7. Healthy

 Enhancements Over Existing System:

• Disease-Specific Classification: Pinpoints individual diseases within broader categories.
• Higher Accuracy: Expected to surpass existing models.
• Improved Generalization: Uses data augmentation and class balancing to avoid overfitting and better handle underrepresented diseases.