Udavith-Reshanjana/vet-diagnostic-models

🐾 Veterinary Diagnostic Models

AI models for multimodal disease diagnosis in domestic animals (cats and dogs).

⚠️ Disclaimer

This is a research prototype for educational and decision-support purposes only.

• Not approved for clinical use
• Always consult qualified veterinarians
• Predictions are suggestions, not diagnoses

📊 Model Description

Model Type: Deep Neural Network
Framework: TensorFlow/Keras
Input: 5 modalities (encounters, biochemistry, hematology, urinalysis, vitals)
Output: 12 disease probabilities

Diseases Detected

• Healthy
• Babesiosis
• Ehrlichiosis
• Chronic Kidney Disease
• Acute Kidney Injury
• Anemia (IMHA)
• Urinary Tract Infection
• Liver Disease
• Diabetes Mellitus
• Thrombocytopenia
• Filariasis (Heartworm)
• Pancreatitis

📁 Files

• 4_all_modalities_model.keras - Trained model
• 4_all_modalities_scaler.pkl - Feature scaler
• 4_all_modalities_features.json - Feature names
• 4_all_modalities_metrics.json - Performance metrics
• metadata.json - Training metadata

🔗 Links

Live Demo: Coming soon on Hugging Face Spaces

👤 Author

Udavith-Reshanjana

🚀 Usage

from huggingface_hub import hf_hub_download
import tensorflow as tf
import pickle

# Download model
model_path = hf_hub_download(
    repo_id="Udavith-Reshanjana/vet-diagnostic-models",
    filename="4_all_modalities_model.keras"
)

# Load model
model = tf.keras.models.load_model(model_path)