arXiv：cs.LG· Xiao Ma, Young D. Kwon, Dong Ma·· 13 小时前
OD-TTA：面向具身视觉系统的按需测试时自适应框架
EmbodiTTA: Resource-Efficient Test-Time Adaptation for Embodied Visual Systems
AI 导读
针对持续测试时自适应（CTTA）在边缘设备上内存与能耗开销过大的问题，研究者提出“按需 TTA”新范式及 OD-TTA 框架，仅在检测到显著域偏移时才触发自适应。OD-TTA 包含轻量域偏移检测、源域模型选择与解耦 BN 更新三项技术，可在小批量下实现内存高效自适应。实验显示其在降低能耗与计算开销的同时取得可比甚至更优的性能。
正文
Abstract:Continual Test-time adaptation (CTTA) continuously adapts the deployed model on every incoming batch of data. While achieving optimal accuracy, existing CTTA approaches present poor real-world applicability on resource-constrained edge devices, due to the substantial memory overhead and energy consumption. In this work, we first introduce a novel paradigm -- on-demand TTA -- which triggers adaptation only when a significant domain shift is detected. Then, we present OD-TTA, an on-demand TTA framework for accurate and efficient adaptation on edge devices. OD-TTA comprises three innovative techniques: 1) a lightweight domain shift detection mechanism to activate TTA only when it is needed, drastically reducing the overall computation overhead, 2) a source domain selection module that chooses an appropriate source model for adaptation, ensuring high and robust accuracy, 3) a decoupled Batch Normalization (BN) update scheme to enable memory-efficient adaptation with small batch sizes. Extensive experiments show that OD-TTA achieves comparable and even better performance while reducing the energy and computation overhead remarkably, making TTA a practical reality.
| Subjects: | Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2505.00986 [cs.LG] |
| (or arXiv:2505.00986v4 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2505.00986 arXiv-issued DOI via DataCite |
Submission history
From: Xiao Ma [view email]
[v1]
Fri, 2 May 2025 04:19:07 UTC (10,904 KB)
[v2]
Sat, 18 Apr 2026 11:34:35 UTC (10,865 KB)
[v3]
Tue, 16 Jun 2026 09:00:29 UTC (21,606 KB)
[v4]
Fri, 9 Oct 2026 03:43:22 UTC (4,690 KB)
来源：arXiv：cs.LG · arxiv.org