Segment objects in images and videos using text prompts