Application Demo
Brief
This project allows children to easily and interestingly recognize and understand the birds they encounter through simple photography.
My Contributions
- Completed market research and PRD output, defined the core process of MVP, and designed a high-fidelity prototype of the "Little Bird Watcher" APP;
- Adopted the ResNet50+LoRA fine-tuning solution, achieved 70% accuracy in the CUB-200 dataset, and built FastAPI backend services;
- Formulated evaluation indicators (accuracy/user satisfaction), and optimized model performance and interactive experience through simulated feedback mechanisms;
Status
Deliver API prototype response time < 2 seconds, complete technical documentation and UI interaction draft, and lay the foundation for subsequent development. Verified an AI recognition model (ResNet50 + LoRA) that achieved ~70% baseline accuracy (CUB validation set).
Click To Play the Demo Video 👉
Detailed Design
