GliTr: Glimpse Transformers with Spatiotemporal Consistency for Online Action Prediction

TL;DR

GliTr introduces a novel glimpse-based transformer model for online action prediction that operates with limited visual information, leveraging spatiotemporal consistency to achieve high accuracy with minimal frame observation.

Contribution

This work proposes GliTr, a glimpse transformer with a spatiotemporal consistency training objective, enabling effective action prediction using only partial frame glimpses.

Findings

Achieves 53.02% accuracy on SSv2 with only 33% of frame area observed.

Outperforms baseline methods by approximately 10% with the proposed consistency objective.

Demonstrates high accuracy (93.91%) on Jester dataset with limited visual input.

Abstract

Many online action prediction models observe complete frames to locate and attend to informative subregions in the frames called glimpses and recognize an ongoing action based on global and local information. However, in applications with constrained resources, an agent may not be able to observe the complete frame, yet must still locate useful glimpses to predict an incomplete action based on local information only. In this paper, we develop Glimpse Transformers (GliTr), which observe only narrow glimpses at all times, thus predicting an ongoing action and the following most informative glimpse location based on the partial spatiotemporal information collected so far. In the absence of a ground truth for the optimal glimpse locations for action recognition, we train GliTr using a novel spatiotemporal consistency objective: We require GliTr to attend to the glimpses with features similar to the corresponding complete frames (i.e. spatial consistency) and the resultant class logits at time $t$ equivalent to the ones predicted using whole frames up to $t$ (i.e. temporal consistency). Inclusion of our proposed consistency objective yields ~10% higher accuracy on the Something-Something-v2 (SSv2) dataset than the baseline cross-entropy objective. Overall, despite observing only ~33% of the total area per frame, GliTr achieves 53.02% and 93.91% accuracy on the SSv2 and Jester datasets, respectively.