ImageNet Large Scale Visual Recognition Challenge

The paper documents the ImageNet Large Scale Visual Recognition Challenge, a benchmark in object category classification and detection spanning hundreds of object categories and millions of images. Run annually from 2010 onward and attracting participation from more than fifty institutions, the paper describes how this large-scale benchmark dataset was created, including the challenges involved in collecting ground truth annotation at this scale.

Beyond describing the dataset's construction, the authors highlight key breakthroughs in categorical object recognition that the challenge enabled, provide a detailed analysis of the state of large-scale image classification and object detection, and compare state-of-the-art computer vision accuracy against human accuracy. The paper concludes with lessons learned across five years of running the challenge and proposes future directions and improvements for the field.