What is the source of images in the MS COCO dataset? Was it manually photographed for the very purpose of creating a dataset through some group of people? or crawled the web? or some image hosting site?
Anybody know? Thanks
They are scraped from Flickr. See the Terms of Use
The COCO Consortium does not own the copyright of the images. Use of the images must abide by the Flickr Terms of Use.