Business Challenge
Manufacturing lines struggle with manual defect detection — slow, error-prone, and costly — while quality teams lack real-time visibility into production anomalies.
Our Solution
Automated visual inspection system using deep learning (YOLO + ResNet) to detect surface defects in real time, reducing false negatives by 94% and cutting QC labor costs by 60%.
CV Inspection Pipeline
- Image Capture — HD cameras @ 60fps
- Preprocessing — noise reduction, normalization
- Defect Detection — YOLOv8 inference <12ms/frame
- Result Output — dashboard alert + auto-reject
Tech Stack
- Python
- PyTorch
- YOLOv8
- OpenCV
- NVIDIA Triton Inference Server
- FastAPI
- Docker
- AWS S3