Skip to content

Vision Configurations

A Vision Configuration defines how a vision system inspects products: which detectors run (top camera, bottom camera, and optional post-processing detectors for defects), which AI model each detector uses, and how captured images are stored. Sorting Recipes are created inside a Vision Configuration and use its detected labels and defects to route products.

A system can have several configurations saved (for example one per product type), but only one can be active on a device at a time.

How Vision Configurations Work

Each configuration can enable up to four detectors:

  • Top Detector — the main detector for the top camera. Enabled by default.
  • Bottom Detector — an optional detector for a bottom camera.
  • Post-Processor Top / Post-Processor Bottom — optional secondary detectors that run after the main detector, typically used for defect detection.

Each detector has its own Task Type:

Task Type Use
Classic Traditional (non-AI) blob/contour detection. Choose this if you want to use Filter Range/Erosion-based calibration, as described in Camera Calibration.
Segmentation AI segmentation that produces pixel-level masks.
Boxes AI object detection using bounding boxes.
Oriented Boxes AI object detection using oriented bounding boxes.
Text Prompts Detects objects described by a text prompt (YOLOE/SAM).
Detect Anything Prompt-free detection (YOLOE/SAM).

Detectors using Segmentation, Boxes, or Oriented Boxes also need an AI Model selected. Classic, Text Prompts, and Detect Anything detectors don't use a selectable model.

Configuration Fields

General

Field Description
Configuration Name The name shown on the configuration's card.
Thumbnail An optional image (jpg/png) shown on the card.
Enable 3D Vision Turns on 3D vision for this configuration, and with it the Options ▸ 3D tab, used to configure the zero plane, point-cloud, and depth-camera alignment.
Use Top/Bottom Detector, Use Post-Processor Top/Bottom Enables the corresponding detector page.

Per detector (Top/Bottom Detector, Post-Processor Top/Bottom)

Field Description
Task Type See the table above.
Tracker Type Not shown for post-processor detectors.
Model Only shown when the Task Type requires a model.
Model Image Size 480, 640, 1024, or 1280.
Score Threshold (%) Minimum AI confidence to count as a detection.
Use Retina Masks Higher accuracy segmentation masks, at the cost of more computation.
Video (for testing) Only on Top/Bottom Detector pages — lets you test against a recorded video instead of a live camera.

Storage

Field Description
Store images in cloud Uploads captured images to the cloud.
Only store images with defects in cloud Only relevant when cloud storage is enabled.
Store images locally Keeps captured images on the device.

Creating a Vision Configuration

  1. Open Configurations from the navigation menu.
  2. Select Add.
  3. On the General page, enter a Configuration Name, optionally upload a Thumbnail, and enable Enable 3D Vision if needed.
  4. Check which detectors you need: Use Top Detector (on by default), Use Bottom Detector, Use Post-Processor Top, Use Post-Processor Bottom.
  5. For each detector you enabled, go to its page and set the Task Type, Tracker Type (not for post-processors), Model (if required), Model Image Size, Score Threshold, and Use Retina Masks. Set a Video (for testing) on the Top/Bottom Detector pages if you want to test without a live camera.
  6. Go to the Storage page and choose whether to store images in the cloud and/or locally.
  7. Select Save.

Editing a Vision Configuration

  1. Open Configurations.
  2. Select the configuration's card, then select Edit.
  3. Update any of the General, detector, or Storage pages.
  4. Select Save.

Activating a Vision Configuration

Only one configuration can be active on a device at a time.

  1. Open Configurations.
  2. Select the configuration you want to use.
  3. Select Activate and confirm.

Activating a configuration also reselects a sorting recipe for it — the one that was last active for that configuration, or its first recipe if it has no previously active one.

Deleting a Vision Configuration

  1. Open Configurations.
  2. Select the configuration's card (it must not be the active one).
  3. Select Delete and confirm.

Calibrating a Configuration

Camera calibration is separate from the configuration editor and is done from Options once a Classic-detector configuration is active. See Camera Calibration.

Note

Adding, editing, and deleting configurations may not be available in every build — if the Add, Edit, or Delete buttons are greyed out for you, check with your system integrator. Activate and uploading configuration files are always available once a vision system is connected.