Industrial Computer Vision Defect Inspection Pipeline (Germany)
AI & Machine Learning Verified Production Architecture

Project Overview

Sub-millimeter automated optical inspection system operating on high-speed automotive assembly lines at 120 frames per second using edge-accelerated deep learning.

Client & Location

Precision Autowerk Engineering AG
Location: Stuttgart, Germany
Industry: Automotive Manufacturing

Challenges

  • Manual human visual inspection missing microscopic micro-cracks in cast components.
  • Assembly line speeds of 120 items per minute requiring sub-10 millisecond inference.
  • Industrial factory floor vibrations and changing lighting conditions.

Solution

Deployed custom-trained YOLOv10 object detection models optimized with NVIDIA TensorRT running on edge NVIDIA Jetson AGX Orin industrial workstations.

Technologies Used

  • PyTorch & YOLOv10 Deep Convolutional Networks
  • NVIDIA TensorRT & DeepStream SDK
  • Industrial GigE Vision Basler Cameras
  • OPC UA Factory Floor Protocol Integration

Results & Business Impact

  • Inspection accuracy reached 99.97% defect detection rate.
  • Defective parts released to downstream assembly reduced to zero.
  • Reduced quality assurance operational overhead by €850,000 annually.
Looking for a similar solution?

Our engineers can build, optimize, and scale this for your business.

Schedule a Call

Project Details

Completed
Domain & Field AI & Machine Learning
Technologies Employed
COMPUTER VISION EDGE AI TENSORRT YOLOV10
Engagement & Delivery Dedicated Agile Pod
US & European Timezone Overlap
Client Confidentiality Bilateral Mutual NDA Signed
Live Deployment https://smsoftwares.com/services/ai-and-machine-learning/data-and-training
Book 30-Min Discovery Call Back to All Projects