Abstract
Gait recognition from video streams is a challenging problem in computer vision biometrics due to the subtle differences between gaits and numerous confounding factors. Recent advancements in self-supervised pretraining have led to the development of robust gait recognition models that are invariant to walking covariates. While neural scaling laws have transformed model development in other domains by linking performance to data, model size, and compute, their applicability to gait remains unexplored. In this work, we conduct the first empirical study scaling on skeleton-based self-supervised gait recognition to quantify the effect of data quantity, model size and compute on downstream gait recognition performance. We pretrain multiple variants of GaitPT - a transformer-based architecture - on a dataset of 2.7 million walking sequences collected in the wild. We evaluate zero-shot performance across four benchmark datasets to derive scaling laws for data, model size, and compute. Our findings demonstrate predictable power-law improvements in performance with increased scale and confirm that data and compute scaling significantly influence downstream accuracy. We further isolate architectural contributions by comparing GaitPT with GaitFormer under controlled compute budgets. These results provide practical insights into resource allocation and performance estimation for real-world gait recognition systems.
Abstract (translated)
视频流中的步态识别是计算机视觉生物特征领域的一项挑战性问题，因为步态之间存在细微差异，并且有许多干扰因素。最近，在自我监督预训练领域的进展已导致开发出一些鲁棒的步态识别模型，这些模型对行走协变量具有不变性。尽管神经规模法则通过将性能与数据量、模型大小和计算能力联系起来，已在其他领域改变了模型的发展方式，但其在步态识别中的适用性尚未被探索。 在这项工作中，我们首次进行了基于骨架的自我监督步态识别的扩展研究，以量化数据量、模型大小和计算资源对下游步态识别性能的影响。我们在一个包含270万条野生采集步行序列的数据集上对GaitPT（一种基于变压器架构）的不同变体进行预训练，并在四个基准数据集上评估零样本性能，从而推导出关于数据量、模型规模和计算能力的扩展法则。 我们的发现表明，在增大规模的情况下，性能会按照可预测的幂律方式得到改进，并确认数据量和计算资源的扩展显著影响下游准确性。我们进一步通过控制计算预算，将GaitPT与GaitFormer进行比较来分离架构贡献。这些结果为实际步态识别系统的资源配置和性能估计提供了实用见解。
URL
https://arxiv.org/abs/2504.07598
