Pashu Dhan AI

Pashu Dhan AI is a comprehensive veterinary diagnostics and decision-support system designed to empower Indian farmers, veterinarians, and researchers. The application integrates computer vision and generative AI to instantly identify 39+ indigenous cattle breeds from images and provide actionable biological insights. Built on Streamlit, it combines specialized local classification models with global large language models for localized agricultural intelligence.

• Automated Breed Classification: Uses a fine-tuned EfficientNetB0 Convolutional Neural Network deployed with TensorFlow to perform real-time image inference with ~91% accuracy across 39 classes.
• Retrieval-Augmented Trait Generation: Queries a localized Pandas-driven CSV database containing physiological and behavioral attributes to build context-enriched prompts for the Google Gemini 3.5 Flash API.
• Dynamic Multilingual Translation: Integrates deep-translator APIs to convert inputs and generated outputs dynamically into regional languages including Hindi, Marathi, and Punjabi.
• On-Demand Text-to-Speech: Employs the gTTS library to synthesize spoken-word audio files from generated reports, caching audio bytes inside the Streamlit session state for low-latency playback.