Javier Lloret Pardo

works
bio
contact
teaching

Annotators' View

Web-based, 2020. Online version here

Image Annotators constitute the hidden labour of AI vision. The current ubiquitous techniques of image classification, segmentation and scene description wouldn't be possible without the manual labour performed by image annotators. They are the ones teaching AI-based computer vision to "see". Recent scene annotation techniques ask annotators to point at what they are looking at while describing the scene in a given image.

"Annotators' View" displays the way image annotators see. It progressively reveals only the areas of the image that annotators' look and point at. The site creates an endless loop by randomly selecting images that have been annotated with this technique from different datasets. These images overwrite each other, creating an ever-changing visual collage.