A New Split for Evaluating True Zero-Shot Action Recognition

TL;DR

This paper introduces a new evaluation split for zero-shot action recognition that prevents class overlap with pre-training data, leading to more realistic and challenging assessments of model performance.

Contribution

It proposes a true zero-shot split avoiding class overlap with pre-training datasets and benchmarks recent methods on this split, highlighting the need for more rigorous evaluation.

Findings

Zero-shot performance drops by up to 8.9% with the new split.

The new split is significantly more challenging than random splits.

Similar issues are found in few-shot splits, with performance differences up to 17.1%.

Abstract

Zero-shot action recognition is the task of classifying action categories that are not available in the training set. In this setting, the standard evaluation protocol is to use existing action recognition datasets(e.g. UCF101) and randomly split the classes into seen and unseen. However, most recent work builds on representations pre-trained on the Kinetics dataset, where classes largely overlap with classes in the zero-shot evaluation datasets. As a result, classes which are supposed to be unseen, are present during supervised pre-training, invalidating the condition of the zero-shot setting. A similar concern was previously noted several years ago for image based zero-shot recognition but has not been considered by the zero-shot action recognition community. In this paper, we propose a new split for true zero-shot action recognition with no overlap between unseen test classes and training or pre-training classes. We benchmark several recent approaches on the proposed True Zero-Shot(TruZe) Split for UCF101 and HMDB51, with zero-shot and generalized zero-shot evaluation. In our extensive analysis, we find that our TruZesplits are significantly harder than comparable random splits as nothing is leaking from pre-training, i.e. unseen performance is consistently lower,up to 8.9% for zero-shot action recognition. In an additional evaluation we also find that similar issues exist in the splits used in few-shot action recognition, here we see differences of up to 17.1%. We publish oursplits1and hope that our benchmark analysis will change how the field is evaluating zero- and few-shot action recognition moving forward.