AI Classification (Multi-Class) Tool

You are currently viewing the documentation for a pre-release version (2.2.0). To access documentation for other versions, click the "Switch Version" button located in the upper-right corner of the page.

■ If you're unsure about the version of the product you are using, please contact Mech-Mind Technical Support for assistance.

Function Introduction

The AI Classification (Multi-Class) Tool is a deep-learning-based visual classification tool supporting custom training for up to 8 classes. Through a wizard workflow, you can quickly complete image capture, labeling, training, and inference deployment. Typical scenarios include mixed-model recognition and appearance-defect grading for a single workpiece family.

overall workflow

  1. Capture Images: Collect representative training data from real production conditions.

  2. Set Target Region: Configure ROI for subsequent labeling and training.

  3. Label Images: Assign class labels to target regions (up to 8 classes).

  4. Train and Validate: Train model, validate performance, and perform additional training if needed.

Workflow

After opening the tool, click New at the upper-right of model list to create and enter a new model workflow.

Capture Images

  1. Ensure input image port is connected.

  2. One image is captured automatically when opening the tool. Click Capture Image to acquire more images.

Include typical variations in data collection:

  • Position and angle changes.

  • Lighting and background changes.

  • Appearance differences such as stains, scratches, or slight deformation.

Set Target Region

If 2D Alignment Parameter Group has input, ROI is transformed with target pose. The configured ROI is the reference ROI.

  1. Click Edit to open target-region settings.

  2. Choose region mode:

Mode Description

Entire Image as Target Region

ROI covers entire image. Suitable when target occupies most of image.

Custom Target Region

Use rectangle or circle tool to draw one ROI accurately around target area.

  1. Optionally set mask region to exclude irrelevant interference.

  2. Click Save and Use to apply settings.

If target and mask layers completely overlap, only the top layer is editable.

Label Images

  1. Click Edit to enter labeling.

  2. Create classes as needed (up to 8 classes) and name them.

  3. Select ROI and click corresponding class Label Image button.

  4. Capture new images and repeat labeling until each class has sufficient samples.

Label diverse data for every class. Avoid ambiguous samples. If production lighting/angle varies, enable augmentation options such as brightness and rotation adaptation.

  1. Click Save and Use after labeling.

Train and Validate Model

  1. Train model with Train.

  2. Validate with Validate.

Parameter Description

Validation Result

Displays predicted class or unknown class.

Confidence

Confidence of current prediction.

Time Cost

Inference time per run (ms).

Confidence Threshold

Minimum confidence for class decision. Below threshold is judged as unknown class.

Default value: 0.5

Heatmap

Visualizes model attention regions when enabled.

Default value: Disabled.

  1. Additional training (optional): Use Additional Training to add misclassified or missing cases, then retrain and revalidate.

  2. Save and apply model: Click Save and Use, then select model in Model Name parameter for inference usage.

Is this page helpful?

You can give a feedback in any of the following ways:

We Value Your Privacy

We use cookies to provide you with the best possible experience on our website. By continuing to use the site, you acknowledge that you agree to the use of cookies. If you decline, a single cookie will be used to ensure you're not tracked or remembered when you visit this website.

  • Preference
  • I accept
  • I don't accept