Problem / context
Manual scouting is time-consuming. This prototype explores automatic extraction of match metrics directly from video.
Design / approach
The system combines a React/Vite frontend, Python/Flask backend, and YOLOv8 detection.
Engineering decisions
The documented architecture separates a React/Vite frontend from a Python/Flask backend and uses YOLOv8 for detection. The source does not specify the model-training, tracking, or metric-calculation design.
Constraints
This was an ambitious scope for a solo build, spanning a web interface, backend, computer vision, and video-derived FRC metrics.
Implementation
React, Vite, Python, Flask, YOLOv8, and computer vision.
Status / results
Prototype. The aim is to track BPS, cycle time, and climb success from match video; validated performance is not documented in the source material.
Lessons learned
It was a good lesson in sizing projects realistically.