AI
Edge AI Vision System for Quality Control
Manufacturing · 2023
Back to ProjectsChallenge
The client needed on-line visual inspection solved within latency, bandwidth, and air-gapped plant constraints. Cloud-only inference was not viable for the line cadence.
Solution
We built an edge vision pipeline with model quantization, camera calibration routines, and operator review UI for uncertain detections. Models stayed on-device; only summarized events left the cell.
Outcome
A pilot cell deployment and model update process were delivered. Defect-rate or throughput claims are omitted pending production validation.
TECH STACK USED
Tools behind this build.
PyTorch
ONNX
Jetson Nano
OpenCV
GALLERY
Project imagery.
Have a project like this in mind?
We take on a small number of projects each quarter, selected for technical depth, not for deal size.
Start a Project