MEEV: Body Mesh Estimation On Egocentric Video

TL;DR

MEEV is a novel method for estimating human body shape and motion from egocentric videos, effectively handling occlusions and blurry images by leveraging multiscale features and pre-training on combined datasets.

Contribution

The paper introduces MEEV, a new approach that improves body mesh estimation from head-mounted camera videos by using multiscale features and dataset pre-training, winning the EgoBody Challenge.

Findings

Achieved 82.30 MPJPE and 92.93 MPVPE on the EgoBody dataset.

Outperformed existing methods in challenging egocentric scenarios.

Demonstrated robustness to occlusions and motion blur.

Abstract

This technical report introduces our solution, MEEV, proposed to the EgoBody Challenge at ECCV 2022. Captured from head-mounted devices, the dataset consists of human body shape and motion of interacting people. The EgoBody dataset has challenges such as occluded body or blurry image. In order to overcome the challenges, MEEV is designed to exploit multiscale features for rich spatial information. Besides, to overcome the limited size of dataset, the model is pre-trained with the dataset aggregated 2D and 3D pose estimation datasets. Achieving 82.30 for MPJPE and 92.93 for MPVPE, MEEV has won the EgoBody Challenge at ECCV 2022, which shows the effectiveness of the proposed method. The code is available at https://github.com/clovaai/meev