Case Study VisionInspect – Automated Edge Computer Vision for Manufacturing (Australia)

Project Overview

VisionInspect is an industrial edge computer vision inspection system built for an advanced materials manufacturer in Melbourne, Australia. The solution performs automated real-time quality control and defect classification along high-speed automated production conveyors.

Client & Location

Precision Fabrication Pty Ltd
Location: Melbourne, VIC, Australia
Industry: Industrial Manufacturing

Challenges

• Microscopic surface cracks and coating defects were missed by manual human visual spot-checks.
• High conveyor belt speed (1.8 meters/second) required instant detection and automated pneumatic rejection.
• Unreliable cloud connectivity on the factory floor necessitated 100% on-premises edge computing.

Solution

SM Softwares deployed an edge-native YOLOv8 object detection model optimized with NVIDIA TensorRT on industrial Jetson hardware, accompanied by high-speed industrial cameras and PLC trigger integration.

Technologies Used

• OpenCV & YOLOv8 Object Detection
• NVIDIA TensorRT & CUDA Optimization
• Python & C++ Embedded Runtime
• Industrial MQTT & Modbus PLC Communication

Results & Business Impact

• 99.6% defect detection accuracy at full production speed.
• Saved an estimated $210,000 AUD annually in post-shipment warranty and recall claims.
• Achieved zero-latency edge inference with under 15ms per frame processing.