When will you do what? - Anticipating Temporal Occurrences of Activities

TL;DR

This paper introduces CNN and RNN-based methods for long-term prediction of future human actions and their durations in videos, addressing a gap in existing short-term focused approaches.

Contribution

It presents novel deep learning models capable of predicting extensive future actions and durations in videos, handling noisy inputs and large action diversity.

Findings

Accurate long-term action predictions demonstrated

Models handle noisy and erroneous input effectively

Effective over videos with many different actions

Abstract

Analyzing human actions in videos has gained increased attention recently. While most works focus on classifying and labeling observed video frames or anticipating the very recent future, making long-term predictions over more than just a few seconds is a task with many practical applications that has not yet been addressed. In this paper, we propose two methods to predict a considerably large amount of future actions and their durations. Both, a CNN and an RNN are trained to learn future video labels based on previously seen content. We show that our methods generate accurate predictions of the future even for long videos with a huge amount of different actions and can even deal with noisy or erroneous input information.