[+]
Getting Started
[-]
Recognition
 [-]
Brain Training Beginner's Guide
   
   
   
   
 [+]
Recognition Documentation
  
  
 [+]
APIs
[+]
Detection
[+]
Segmentation
[+]
Resources
Updated on 9/11/2019
Brain Builder Knowledge Base
Walk-Through 1: Training your First Brain
Direct link to topic in this publication:
  • Recognition
  • »
  • Brain Training Beginner's Guide
  • »
  • Walk-Through 1: Training your First Brain

In This Section:

In This Walk-Through:

  • Train your first Brain
  • Understand the tagging/training process
  • See Brain Predictions
  • Time: 5-10 minutes

AI in Sheet Metal Inspection

There are a lot of technologies that inspect products on assembly lines, looking for defects and manufacturing errors. Many of these applications don't use AI because the consistency of the assembly line means simpler technologies can find defects easily enough.

However, some use cases are too complex for these legacy technologies. If the presentation of the products or defects is too varied, an inflexible solution can't adapt or generalize to the range of what it might need to see. That's where a Brain Builder-trained AI Brain can shine.

The Challenge

You are training a Brain to distinguish between two types of defects in a sheet metal production line. These two defect types have distinct appearances, though they vary from image to image.

On the production line, the camera is in a fixed position taking consistent images of the sheet metal as it progresses down the line.

For our needs today, we'll refer to these defect types as Defect A and Defect B.

Defect A Defect B

Data Source and Citation: K. Song and Y. Yan, “A noise robust method based on completed local binary patterns for hot-rolled steel strip surface defects,” Applied Surface Science, vol. 285, pp. 858-864, Nov. 2013.(paper) http://faculty.neu.edu.cn/yunhyan/NEU_surface_defect_database.html

Train your Brain

  1. Log into Brain Builder, and click on the Getting Started project. You'll see multiple Datasets, one of which is labeled 1. Sheet Metal Inspection.

    You can see that this Dataset has 20 images.

  2. Click TAG to start training your Brain. This will take you into the Workspace for the first image in the Dataset.

    This Dataset already has the Classes defined and listed on the right side of the screen in the Classes panel.

  3. Click on the correct label in the Classes panel. (For example, the image above shows Defect A, so you would click on Defect A and save it. You may see a different image depending on how the images are sorted in your Dataset. Use the two images up above as a reference to determine which label to apply.)

    When you save the label, your Brain will learn it and progress to the next image.

  4. Continue labeling images. You'll soon notice that the Brain starts to make Predictions.

    The predictions are the result of the Brain applying the knowledge it has learned to the new image.

    To the right of each prediction is a strength indicator. More green squares indicate a stronger prediction, while fewer squares mean the Brain is less confident in the accuracy of its prediction.

    When Brain Builder makes a prediction, you can click on the prediction instead of the class label to save that knowledge into the Brain.

Read more in Training your Brain.

Wrap-Up

After this walk-through, you should feel comfortable:

  • Navigating the Brain Builder Recognition workspace
  • Training a simple Brain with a small number of images
  • Reading and evaluating the Brain's predictions in the workspace

Next: Walk-Through 2: Getting Smarter >>