Are we done with ImageNet?

TL;DR

This paper questions whether recent improvements on ImageNet truly reflect better generalization or if models are overfitting to its labeling quirks, by developing a more robust annotation method and reevaluating classifier performance.

Contribution

The authors introduce a more reliable annotation process for ImageNet validation data and reassess classifier accuracy, revealing smaller gains and diminishing label predictiveness.

Findings

Recent classifier gains are smaller with new labels.

Original ImageNet labels are less predictive of independent annotations.

Improved annotations fix many labeling errors.

Abstract

Yes, and no. We ask whether recent progress on the ImageNet classification benchmark continues to represent meaningful generalization, or whether the community has started to overfit to the idiosyncrasies of its labeling procedure. We therefore develop a significantly more robust procedure for collecting human annotations of the ImageNet validation set. Using these new labels, we reassess the accuracy of recently proposed ImageNet classifiers, and find their gains to be substantially smaller than those reported on the original labels. Furthermore, we find the original ImageNet labels to no longer be the best predictors of this independently-collected set, indicating that their usefulness in evaluating vision models may be nearing an end. Nevertheless, we find our annotation procedure to have largely remedied the errors in the original labels, reinforcing ImageNet as a powerful benchmark for future research in visual recognition.