Back to Projects

Edge Vision Quality Control

Ultralytics YOLO + Docker on Raspberry Pi to inspect packaging in real time.

Computer VisionYOLODockerRaspberry Pi
<120ms inference on Raspberry Pi

Context & Problem

Addressed quality gaps where manual oversight failed to identify micro-defects during night shifts, resulting in costly inefficiencies and material waste.

Approach

Trained YOLO models on labeled batches and optimized packaging images.

Architecture

Camera → Raspberry Pi running Docker → YOLO inference.

Challenges

Kept inference under 120ms despite limited GPU resources by tuning model sizes and leveraging TensorRT conversions.

Tech Stack

Computer VisionYOLODockerRaspberry Pi