Projects - ML systems

Yepai ML Route Grader

ML route grading for climbing walls.

20251st placeYOLOv8PyTorchGNNsAWS EC2Next.js

Problem

Indoor climbing grades are subjective, and computer vision has to infer route structure from limited visual data and sparse labels.

Built dual YOLOv8 models for hold detection and color classification with a limited dataset.
Implemented a graph neural network connecting reachable holds with hold-type features.
Deployed the model on AWS EC2 and integrated it with a Next.js app for real-time predictions.

Edgar Yepez