ObjectMix: Data Augmentation by Copy-Pasting Objects in Videos for Action Recognition

TL;DR

ObjectMix is a novel data augmentation technique for action recognition that combines object regions from different videos using instance segmentation, improving performance on UCF101 and HMDB51 datasets.

Contribution

It introduces ObjectMix, a new method that leverages instance segmentation to create augmented videos by combining objects from different videos, tailored for action recognition.

Findings

Outperforms VideoMix in experiments

Effective on UCF101 and HMDB51 datasets

Enhances action recognition accuracy

Abstract

In this paper, we propose a data augmentation method for action recognition using instance segmentation. Although many data augmentation methods have been proposed for image recognition, few of them are tailored for action recognition. Our proposed method, ObjectMix, extracts each object region from two videos using instance segmentation and combines them to create new videos. Experiments on two action recognition datasets, UCF101 and HMDB51, demonstrate the effectiveness of the proposed method and show its superiority over VideoMix, a prior work.