The Flamingo project aims to conduct exploratory research through crude semantic segmentation and assigning a variety of words to images in order to create a "dumb agent" that can provide information on the image with its limited knowledge base.