This project is an exploration into the capabilities of fine-tuning the Llava model specifically for Indian desserts. We've successfully fine-tuned the model to identify and provide in-depth nutritional information on 5 distinct Indian dessert items. This not only includes a brief overview of each dessert, covering aspects like taste and nutritional content, but also allergen notes, recommended times to eat, and feel after consumption.
Watch our model in action and see how it beautifully integrates AI with culinary tradition:
🌟 Key Highlights:
- Enhanced Detection: Our model can recognize some dessert items that were previously undetectable by standard Llava and even GPT-4 models.
- Image Recognition: It showcases improved accuracy in image detection, surpassing GPT-4v in recognizing specific images and providing detailed information.
- Replicate Training: Utilized Replicate for training and hosting, ensuring a seamless integration and user experience.
Experience the model firsthand and explore its capabilities: Llava Fine-Tuned Model on Replicate
This project serves as a dynamic showcase of how fine-tuning LVM (Language Vision Models) on domain-specific datasets can significantly enhance AI's understanding and analysis capabilities.
- PCB Verification: Training LVMs on images of PCBs (Printed Circuit Boards) to verify and filter them based on specific queries.
- Domain-Specific Training: Tailoring the model to recognize and evaluate domain-specific or private image datasets for more effective and accurate recognition.
To get started with this project and explore the fine-tuned Llava model's capabilities, please follow the links provided above. Your journey into the advanced application of AI in understanding and providing detailed insights into Indian desserts begins here!
Contributions, questions, and feedback are welcome! Feel free to open an issue or submit a pull request.