AI-Powered Fungi Detection

A hybrid deep learning system for fungal species classification, developed as the lead project for the Olympic AI HCMC Competition where it earned 4th place.

Designed a hybrid ConvNeXt + Transformer model that combines the local feature extraction strength of ConvNeXt with the global attention mechanism of Transformers, achieving 97.7% accuracy in fungal species classification.

Automated the end-to-end pipeline: data ingestion, cleaning, augmentation, and preprocessing. Led a 4-member team, coordinated research direction, and prepared presentation materials for the competition judges.