Mimetics: Towards Understanding Human Actions Out of Context

TL;DR

This paper introduces Mimetics, a new dataset for evaluating human action recognition without contextual cues, revealing that current models struggle without context and that pose-based methods are more robust.

Contribution

The paper presents Mimetics, a novel dataset for out-of-context action recognition, and demonstrates the effectiveness of pose-based models over traditional 3D CNNs in this setting.

Findings

State-of-the-art 3D CNNs perform poorly on mimed actions without context.

Pose-based models are less biased by scene or object context.

A simple neural network on pose features performs competitively.

Abstract

Recent methods for video action recognition have reached outstanding performances on existing benchmarks. However, they tend to leverage context such as scenes or objects instead of focusing on understanding the human action itself. For instance, a tennis field leads to the prediction playing tennis irrespectively of the actions performed in the video. In contrast, humans have a more complete understanding of actions and can recognize them without context. The best example of out-of-context actions are mimes, that people can typically recognize despite missing relevant objects and scenes. In this paper, we propose to benchmark action recognition methods in such absence of context and introduce a novel dataset, Mimetics, consisting of mimed actions for a subset of 50 classes from the Kinetics benchmark. Our experiments show that (a) state-of-the-art 3D convolutional neural networks obtain disappointing results on such videos, highlighting the lack of true understanding of the human actions and (b) models leveraging body language via human pose are less prone to context biases. In particular, we show that applying a shallow neural network with a single temporal convolution over body pose features transferred to the action recognition problem performs surprisingly well compared to 3D action recognition methods.