Knowledge Distillation for Human Action Anticipation

Abstract

We consider the task of training a neural network to anticipate human actions in video. This task is challenging given the complexity of video data, the stochastic nature of the future, and the limited amount of annotated training data. In this paper, we propose a novel knowledge distillation framework that uses an action recognition network to supervise the training of an action anticipation network, guiding the latter to attend to the relevant information needed for correctly anticipating the future actions. This framework is possible thanks to a novel loss function to account for positional shifts of semantic concepts in a dynamic video. The knowledge distillation framework is a form of self-supervised learning, and it takes advantage of unlabeled data. Experimental results on JHMDB and EPIC-KITCHENS dataset show the effectiveness of our approach.

Tragedy Plus Time: Capturing Unintended Human Activities from Weakly-labeled Videos

“…[22,24,39,48,49] focus on forecasting an event or action based on a small snippet of a video. [62,67] focus on self supervised learning approaches to predict future action representation using unlabeled videos. [20,47,63] focus on predicting a pedestrian's intent to cross the road using the Joint Attention for Autonomous Driving (JAAD) dataset [46].…”