BLAPose: Enhancing 3D Human Pose Estimation with Bone Length Adjustment

TL;DR

This paper introduces a bone length adjustment technique for 3D human pose estimation that enforces physical constraints, improving model accuracy by predicting and refining bone lengths within video sequences.

Contribution

The work presents a novel recurrent neural network architecture, synthetic bone length augmentation, and a bone length adjustment method that collectively enhance 3D human pose estimation accuracy.

Findings

Bone length prediction surpasses previous best results.

Adjustment method improves existing pose estimation models.

Fine-tuning with inferred bone lengths yields notable performance gains.

Abstract

Current approaches in 3D human pose estimation primarily focus on regressing 3D joint locations, often neglecting critical physical constraints such as bone length consistency and body symmetry. This work introduces a recurrent neural network architecture designed to capture holistic information across entire video sequences, enabling accurate prediction of bone lengths. To enhance training effectiveness, we propose a novel augmentation strategy using synthetic bone lengths that adhere to physical constraints. Moreover, we present a bone length adjustment method that preserves bone orientations while substituting bone lengths with predicted values. Our results demonstrate that existing 3D human pose estimation models can be significantly enhanced through this adjustment process. Furthermore, we fine-tune human pose estimation models using inferred bone lengths, observing notable improvements. Our bone length prediction model surpasses the previous best results, and our adjustment and fine-tuning method enhance performance across several metrics on the Human3.6M dataset.