UV R-CNN: Stable and Efficient Dense Human Pose Estimation

TL;DR

UV R-CNN introduces a novel architecture with a new loss function and balanced training strategy, significantly improving dense human pose estimation accuracy without auxiliary supervision.

Contribution

The paper proposes UV R-CNN with a Dense Points loss and balanced loss weighting, enhancing training stability and performance in dense human pose estimation.

Findings

Achieves 65.0% AP_{gps} on DensePose-COCO validation.

Outperforms previous methods in dense human pose estimation.

Stable training process without auxiliary supervision.

Abstract

Dense pose estimation is a dense 3D prediction task for instance-level human analysis, aiming to map human pixels from an RGB image to a 3D surface of the human body. Due to a large amount of surface point regression, the training process appears to be easy to collapse compared to other region-based human instance analyzing tasks. By analyzing the loss formulation of the existing dense pose estimation model, we introduce a novel point regression loss function, named Dense Points} loss to stable the training progress, and a new balanced loss weighting strategy to handle the multi-task losses. With the above novelties, we propose a brand new architecture, named UV R-CNN. Without auxiliary supervision and external knowledge from other tasks, UV R-CNN can handle many complicated issues in dense pose model training progress, achieving 65.0% $AP_{gps}$ and 66.1% $AP_{gpsm}$ on the DensePose-COCO validation subset with ResNet-50-FPN feature extractor, competitive among the state-of-the-art dense human pose estimation methods.