Manufacturing Defect Detection Using Computer Vision

Use Object Detection techniques to quickly identify and locate product flaws.

Enhanced Product Quality and Reliability

Employing advanced machine learning for defect detection ensures higher-quality products, meets industry standards, and enhances overall reliability.

Operational Efficiency and Cost Reduction

Automating defect identification in industrial processing with computer vision optimizes efficiency, reduces production costs, and enables more effective resource allocation.

Competitive Edge through Innovation

Adopting machine learning-powered solutions streamlines development, positioning manufacturers as innovators with a competitive edge in delivering top-notch products for wide-ranging markets.

How It Works

In this solution, we leverage computer vision to identify product defects in hot-rolled steel plates, widely used in construction and agriculture. Using an object detection model powered by machine learning, we achieve precise and efficient detection and classification of prevalent defects like scratches.

Traditionally, the visual inspection of steel plates is time-consuming and potentially unreliable. Our approach automates the detection process, enhancing accuracy and reducing human effort and error.

This solution is split into two phases, with model training contained in Phase 1 and model deployment with DataRobot detailed in Phase 2.

Phase 1 uses a dataset containing images of steel plates with different defects to fine-tune a pre-trained Faster R-CNN model, an incredibly powerful object detection machine learning architecture, to classify and locate scratches. 

 How Faster RCNN works

Phase 2 uses the DataRobot platform to host a non-DataRobot, pre-trained object detection model to leverage a host of ML Production features such as model monitoring with custom metrics. The specific code applied here assumes the input/output schema of the resulting FasterR-CNN model from Phase 1 but any object detection model can work with the proper data handling updates.

The output: a highly accurate and robust predictive model capable of detecting and classifying any sized scratch present in steel plates, with a DataRobot deployment monitoring model health and logging metrics. 

image 1
 The model detects a scratch

Key Deliverables

  • AI-based defect detection solution
  • Demonstration of data acquirement and preprocessing 
  • Custom training and validation loops for fine-tuning the model
  • Significantly automated defect detection process
  • Model observability in production with custom metrics to evaluate performance
  • Practical application insights to other manufacturing industries

About the Partner

KUNGFU.AI is a management consulting and engineering firm focused exclusively on artificial intelligence. With a deep understanding of key contributors to AI outcomes, including data, designers, users, and leaders, KUNGFU.AI tailors strategies that propel businesses forward. 

The team, comprising experts in AI, machine learning, and software engineering, stays at the forefront of research to deliver cutting-edge solutions. By truly understanding client challenges and environments, KUNGFU.AI iteratively crafts solutions, optimizing AI operations at every stage. Their commitment extends beyond implementation to repeatability, reliability, and alignment with corporate objectives, empowering organizations to redirect resources toward innovation and sustainable advantage. 

Whether clients are initiating a data strategy or deploying production AI models, KUNGFU.AI ensures maximum return on investment in an AI-driven future.


Solution Architecture

Defect Detection Using Computer Vision Solution Architecture

Success Story

KUNGFU.AI designed and developed a computer vision detection and pricing system capable of detecting dents of various sizes and magnitudes to calculate price estimates for repairs.

The client, an automotive services company, sought innovative ways to automate service delivery and expand offerings through tech-enabled solutions. Our objective was to create a capability enabling customers to send images for automated damage inspection and cost estimates for common car repairs.

This capability provides the client with a new source of revenue while reducing delivery overhead. It has initiated channel distribution alliances with third-party auto-traders and attributed to valuation expansion for acquisition by Repairify.

Get Started with Free Trial

Experience new features and capabilities previously only available in our full AI Platform product.

Related AI Accelerator
Steel Plate Defect Object Detection

In this notebook, we are going to focus on detecting and classifying product defects using state of the art computer vision systems.

View Accelerator  

Explore the Most Popular Resources

cta module 1920px

Take AI From Vision to Value

See how a value-driven approach to AI can accelerate time to impact.