TrueBreed Identifier

• 

  this response is generated from ml model after analysing images.

Inspiration Farmers, vets, and researchers in rural areas face challenges identifying cattle breeds accurately, which affects livestock management, insurance claims, and breed conservation. We wanted to build a solution that not only identifies breeds using AI but also educates users on breed specifics, empowering better decisions and protecting farmers from fraud.

What it does TrueBreed Identifier recognizes cattle and animal breeds from photos using machine learning and provides breed-specific information on demand. It combines image recognition with text feature extraction and integrates Google’s Large Language Model API to answer user questions about breed characteristics and care.

How we built it Cleaned and preprocessed datasets with breed and catalog info using regex for pack quantity extraction.

Applied TF-IDF vectorization for text features merged with numeric data in a LightGBM regression model for breed prediction.

Developed a Flutter frontend and Flask backend for user interaction and API serving.

Integrated Google Gemini LLM API to respond to user queries in natural language about breeds and cattle specifications.

Challenges we ran into Dealing with inconsistent and noisy real-world data including varied catalog descriptions and limited labeled images.

Ensuring backend and frontend worked smoothly together, overcoming CORS and deployment issues.

Balancing response speed and prediction accuracy to provide instant user feedback even on low-resource devices.

Fine-tuning the LLM prompt engineering for concise, context-aware breed knowledge responses.

Accomplishments that we're proud of Successfully created an end-to-end AI system combining computer vision, NLP, and large language models.

Built a real-time, interactive app usable by non-experts for breed recognition and education.

Integrated Google’s LLM API, enabling a virtual livestock assistant capable of answering detailed breed-related questions.

What we learned How to preprocess complex multi-modal datasets for robust AI modeling.

The power of combining traditional ML models with generative AI for enhanced user experience.

Techniques for scaling and deploying ML solutions on cloud and mobile platforms with real-time performance.

What's next for TrueBreed Identifier Extend breed database for more species and regional varieties.

Implement dynamic learning to adapt to new breeds and user feedback.

Add voice-based Q&A using LLM for hands-free accessibility.

Integrate IoT sensors to combine visual data with health/environmental parameters for holistic livestock monitoring.

Built With

• flaskapi
• flutter
• mysql