PE-former: Pose Estimation Transformer

TL;DR

This paper introduces PE-former, a pure transformer architecture for 2D body pose estimation, achieving state-of-the-art results without using CNN backbones, demonstrating transformers' effectiveness in complex vision tasks.

Contribution

It presents the first pure transformer approach for 2D pose estimation, eliminating the need for convolutional backbones and setting new performance benchmarks.

Findings

Pure transformer architecture achieves state-of-the-art results.

Encoder-decoder design is effective for pose estimation.

Transformer-based models outperform CNN-based methods.

Abstract

Vision transformer architectures have been demonstrated to work very effectively for image classification tasks. Efforts to solve more challenging vision tasks with transformers rely on convolutional backbones for feature extraction. In this paper we investigate the use of a pure transformer architecture (i.e., one with no CNN backbone) for the problem of 2D body pose estimation. We evaluate two ViT architectures on the COCO dataset. We demonstrate that using an encoder-decoder transformer architecture yields state of the art results on this estimation problem.