AI segmentation

Note

The following requirements are needed for enabling this feature:

• On the hardware side, a standard or high performance Pickit processor, which have a dedicated GPU.

• On the software side, the AI segmentation license module.

AI segmentation uses a Visual-Language Model (VLM) to identify entire categories of objects using simple text prompts. It’s used together with detection engines like Pickit Flex or Pickit Pattern.

The prompt determines the object category to detect, and the threshold determines how strict the clustering is. Higher threshold values result in stricter clustering. This value is application dependent; some applications need lower values and work reliably, while others might need higher values.

When AI segmentation is toggled off, clustering presets are used for grouping points into clusters.

An example use case for AI segmentation is detecting mixed boxes in a bin picking application (shown below, bottom-left): the prompt box can be used with Pickit Flex’s variable size rectangle detection. This combination eliminates the need for modeling individual SKUs, enabling applications with a high-mix of SKUs. Some example applications are shown below.