Computer-Vision Food Quality & Plating Consistency Checker for a Multi-Location Restaurant Brand

Overview

Built a computer-vision quality-check tool for a multi-location restaurant brand concerned that signature dishes weren't looking consistent across different kitchens and shifts. Line cooks photograph a finished plate at pass, and the system compares it against the brand's reference presentation standard, flagging noticeable deviations in portioning or plating for a manager to review before the dish goes out. This gave brand leadership an objective way to monitor plating consistency across locations instead of relying on periodic in-person visits that only caught a small sample of service. The tool was scoped as a quality-assurance aid for managers, not a replacement for chef judgment on the line. A multi-location restaurant brand's signature dishes weren't looking consistent across different kitchens and shifts, and periodic in-person visits only caught a small sample of service, missing most instances of plating drift. We built a computer-vision tool where line cooks photograph a finished plate at pass, comparing it against the brand's reference presentation standard and flagging noticeable deviations for a manager to review before the dish goes out. We built the reference presentation standards directly from the brand's own signature-dish photography, then tuned deviation thresholds with head chef input to distinguish real drift from acceptable variation. It launched at a few locations first to calibrate sensitivity before expanding brand-wide. Brand leadership now has visibility into plating consistency across every location for the first time, catching drift from brand standards before customers ever noticed it.