SMV-EAR: Bring Spatiotemporal Multi-View Representation Learning into Efficient Event-Based Action Recognition

TL;DR

This paper introduces SMV-EAR, a novel spatiotemporal multi-view representation learning framework for event-based action recognition that improves accuracy and efficiency by addressing limitations of previous methods.

Contribution

It proposes a translation-invariant dense event conversion, a dual-branch fusion architecture, and a bio-inspired temporal augmentation, advancing the state-of-the-art in event-based action recognition.

Findings

Achieved over 7-10% accuracy improvements on three datasets.

Reduced model parameters by 30.1% and computations by 35.7%.

Established a new effective paradigm for event-based action recognition.

Abstract

Event cameras action recognition (EAR) offers compelling privacy-protecting and efficiency advantages, where temporal motion dynamics is of great importance. Existing spatiotemporal multi-view representation learning (SMVRL) methods for event-based object recognition (EOR) offer promising solutions by projecting H-W-T events along spatial axis H and W, yet are limited by its translation-variant spatial binning representation and naive early concatenation fusion architecture. This paper reexamines the key SMVRL design stages for EAR and propose: (i) a principled spatiotemporal multi-view representation through translation-invariant dense conversion of sparse events, (ii) a dual-branch, dynamic fusion architecture that models sample-wise complementarity between motion features from different views, and (iii) a bio-inspired temporal warping augmentation that mimics speed variability of real-world human actions. On three challenging EAR datasets of HARDVS, DailyDVS-200 and THU-EACT-50-CHL, we show +7.0%, +10.7%, and +10.2% Top-1 accuracy gains over existing SMVRL EOR method with surprising 30.1% reduced parameters and 35.7% lower computations, establishing our framework as a novel and powerful EAR paradigm.