Portfolio

Depth Estimation

We solved the notoriously difficult ground truth collection problem in depth estimation for autonomous forklifts by developing a complete data infrastructure with custom annotation tools and a two-stage self-supervised pipeline. Our solution processed 40,000+ images and delivered robust depth datasets that powered accurate real-time autonomous navigation.

Lab Equipement

We transformed a zero-data scenario into a production-ready iOS object detection system in just 10 weeks, building the complete data pipeline from web scraping and custom annotation tools to dataset deployment. Our team managed the entire data lifecycle, including a three-person annotation team, delivering high-quality annotated datasets that powered successful computer vision deployment.

AprilTag Detector

We delivered a complete visual marker detection system for robotics and logistics, combining synthetic data generation with custom detection algorithms to solve real-world reliability issues. The client confirmed our solution outperformed all previous implementations with measurably higher accuracy in production environments.

Badminton Shuttlecock Tracking

We enhanced an award-winning badminton analytics platform through comprehensive dataset analysis, strategic augmentation, and custom post-processing pipelines that improved tracking accuracy during high-velocity gameplay. Our data-driven optimization delivered measurably higher accuracy and reliability for professional athlete use.