Large-Scale Open-Set Classification Protocols for ImageNet

TL;DR

This paper introduces three realistic open-set classification protocols using ImageNet subsets, along with a new validation metric, to evaluate deep learning models' ability to classify known and reject unknown images in real-world scenarios.

Contribution

It proposes new open-set protocols based on ImageNet for more realistic evaluation and introduces a validation metric to assess both classification and rejection performance.

Findings

Algorithms perform well on known negative samples.

Partial success in out-of-distribution detection.

Performance drops with unseen unknown classes.

Abstract

Open-Set Classification (OSC) intends to adapt closed-set classification models to real-world scenarios, where the classifier must correctly label samples of known classes while rejecting previously unseen unknown samples. Only recently, research started to investigate on algorithms that are able to handle these unknown samples correctly. Some of these approaches address OSC by including into the training set negative samples that a classifier learns to reject, expecting that these data increase the robustness of the classifier on unknown classes. Most of these approaches are evaluated on small-scale and low-resolution image datasets like MNIST, SVHN or CIFAR, which makes it difficult to assess their applicability to the real world, and to compare them among each other. We propose three open-set protocols that provide rich datasets of natural images with different levels of similarity between known and unknown classes. The protocols consist of subsets of ImageNet classes selected to provide training and testing data closer to real-world scenarios. Additionally, we propose a new validation metric that can be employed to assess whether the training of deep learning models addresses both the classification of known samples and the rejection of unknown samples. We use the protocols to compare the performance of two baseline open-set algorithms to the standard SoftMax baseline and find that the algorithms work well on negative samples that have been seen during training, and partially on out-of-distribution detection tasks, but drop performance in the presence of samples from previously unseen unknown classes.