Rating is available when the video has been rented.
This feature is not available right now. Please try again later.
Published on Jun 29, 2015
This video shows 80,000 training images from the Microsoft Common Objects in Context (MS COCO) dataset. The order of the images is determined by a meandering walk through a space in which semantically similar images are close together. Specifically, we use the fc7 features fine-tuned on image captions (From Captions to Visual Concepts and Back, Fang et al., CVPR 2015). The middle illustration represents the fc7 features. Higher activations have longer "rays".
For more information on the images and the MS COCO dataset, please visit http://mscoco.org .