ReMA: A Training-Free Plug-and-Play Mixing Augmentation for Video Behavior Recognition

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Cui, Feng-Qi, Huang, Jinyang, Zhao, Sirui, Guo, Jinglong, Cai, Qifan, Yan, Xin, Liu, Zhi
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918268949757952
author Cui, Feng-Qi
Huang, Jinyang
Zhao, Sirui
Guo, Jinglong
Cai, Qifan
Yan, Xin
Liu, Zhi
author_facet Cui, Feng-Qi
Huang, Jinyang
Zhao, Sirui
Guo, Jinglong
Cai, Qifan
Yan, Xin
Liu, Zhi
contents Video behavior recognition demands stable and discriminative representations under complex spatiotemporal variations. However, prevailing data augmentation strategies for videos remain largely perturbation-driven, often introducing uncontrolled variations that amplify non-discriminative factors, which finally weaken intra-class distributional structure and representation drift with inconsistent gains across temporal scales. To address these problems, we propose Representation-aware Mixing Augmentation (ReMA), a plug-and-play augmentation strategy that formulates mixing as a controlled replacement process to expand representations while preserving class-conditional stability. ReMA integrates two complementary mechanisms. Firstly, the Representation Alignment Mechanism (RAM) performs structured intra-class mixing under distributional alignment constraints, suppressing irrelevant intra-class drift while enhancing statistical reliability. Then, the Dynamic Selection Mechanism (DSM) generates motion-aware spatiotemporal masks to localize perturbations, guiding them away from discrimination-sensitive regions and promoting temporal coherence. By jointly controlling how and where mixing is applied, ReMA improves representation robustness without additional supervision or trainable parameters. Extensive experiments on diverse video behavior benchmarks demonstrate that ReMA consistently enhances generalization and robustness across different spatiotemporal granularities.
format Preprint
id arxiv_https___arxiv_org_abs_2601_00311
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle ReMA: A Training-Free Plug-and-Play Mixing Augmentation for Video Behavior Recognition
Cui, Feng-Qi
Huang, Jinyang
Zhao, Sirui
Guo, Jinglong
Cai, Qifan
Yan, Xin
Liu, Zhi
Computer Vision and Pattern Recognition
Video behavior recognition demands stable and discriminative representations under complex spatiotemporal variations. However, prevailing data augmentation strategies for videos remain largely perturbation-driven, often introducing uncontrolled variations that amplify non-discriminative factors, which finally weaken intra-class distributional structure and representation drift with inconsistent gains across temporal scales. To address these problems, we propose Representation-aware Mixing Augmentation (ReMA), a plug-and-play augmentation strategy that formulates mixing as a controlled replacement process to expand representations while preserving class-conditional stability. ReMA integrates two complementary mechanisms. Firstly, the Representation Alignment Mechanism (RAM) performs structured intra-class mixing under distributional alignment constraints, suppressing irrelevant intra-class drift while enhancing statistical reliability. Then, the Dynamic Selection Mechanism (DSM) generates motion-aware spatiotemporal masks to localize perturbations, guiding them away from discrimination-sensitive regions and promoting temporal coherence. By jointly controlling how and where mixing is applied, ReMA improves representation robustness without additional supervision or trainable parameters. Extensive experiments on diverse video behavior benchmarks demonstrate that ReMA consistently enhances generalization and robustness across different spatiotemporal granularities.
title ReMA: A Training-Free Plug-and-Play Mixing Augmentation for Video Behavior Recognition
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2601.00311