HandyPose: Multi-level framework for hand pose estimation

作者：

Highlights：

• We propose HandyPose, a multi-level and multi-scale, end-to-end train-able, singlestage framework for 2D hand pose estimation.

• We introduce the Multi-level Waterfall Atrous Spatial Pooling module that effectively encodes feature maps with large FOV and contextual information.

• HandyPose is a modular encoder-decoder architecture that incorpo-rates multilevel features in both the encoder and the decoder modules,making it easy to modify and expand.

• HandyPose achieves state-of-the-art result for 2D hand pose on two popular benchmarks.

摘要

•We propose HandyPose, a multi-level and multi-scale, end-to-end train-able, singlestage framework for 2D hand pose estimation.•We introduce the Multi-level Waterfall Atrous Spatial Pooling module that effectively encodes feature maps with large FOV and contextual information.•HandyPose is a modular encoder-decoder architecture that incorpo-rates multilevel features in both the encoder and the decoder modules,making it easy to modify and expand.•HandyPose achieves state-of-the-art result for 2D hand pose on two popular benchmarks.

论文关键词：Hand pose estimation,Feature representations,Computer vision

论文评审过程：Received 13 July 2021, Revised 7 February 2022, Accepted 27 March 2022, Available online 1 April 2022, Version of Record 7 April 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108674