Simon Willison's blog post links to Google's PaliGemma model, an openly licensed Vision Language Model (VLM) that can identify and outline objects in an image. via @simonw