TempoFormer: A Transformer for Temporally-aware Representations in Change Detection

TL;DR

TempoFormer is a novel transformer-based model that effectively captures temporal dynamics in change detection tasks, outperforming previous methods by integrating a new temporal rotary positional embedding.

Contribution

It introduces the first task-agnostic, temporally-aware transformer architecture with a novel temporal rotary positional embedding for dynamic representation learning.

Findings

Achieved state-of-the-art results on three real-time change detection benchmarks.

Demonstrated flexibility of the architecture across different transformer models.

Outperformed recurrent and temporally-agnostic approaches in accuracy and efficiency.

Abstract

Dynamic representation learning plays a pivotal role in understanding the evolution of linguistic content over time. On this front both context and time dynamics as well as their interplay are of prime importance. Current approaches model context via pre-trained representations, which are typically temporally agnostic. Previous work on modelling context and temporal dynamics has used recurrent methods, which are slow and prone to overfitting. Here we introduce TempoFormer, the first task-agnostic transformer-based and temporally-aware model for dynamic representation learning. Our approach is jointly trained on inter and intra context dynamics and introduces a novel temporal variation of rotary positional embeddings. The architecture is flexible and can be used as the temporal representation foundation of other models or applied to different transformer-based architectures. We show new SOTA performance on three different real-time change detection tasks.