Class: 232 (Border Collie), Feature: 1065

MTurk Worker Annotations

We now show the responses from five crowdworkers used to annotate this class-feature pair as core or spurious. Workers first denote if the focus of the image is on the main object (core), or a separate or background object (spurious). They also mark their confidence on a scale of 1 to 5, and provide written explanations for their annotation.



Main Focus	Confidence	Explanation
Main Object	3	Focus on the main object
Main Object	5	The focus of the red region in the highlighted visual attributes is on the wood on the dog.
Main Object	4	The red region mainly focuses on the predicted object but slightly spreads to the background.
Main Object	5	The focus is on the muzzle of the border collie.
Main Object	4	Most of the images focusing on some part of the main object


Votes for core: 5, Votes for spurious: 0. Feature is thus annotated as core for the class.