Overview
Who this competition fits
适合已有 Python 深度学习实践基础、愿意使用 PaddlePaddle 2.0 及以上版本的个人参赛者,尤其适合想学习基于骨骼点时序数据进行细粒度人体动作识别的人。参赛者需要能处理由 OpenPose 提取并以 .npy 保存的逐帧骨骼点数据。
Read the original official blurb
2021 CCF BDCI 基于飞桨实现花样滑冰选手骨骼点动作识别赛题再开放,旨在探索基于骨骼点的细粒度人体动作识别方法,要求参赛选手构建基于骨骼点的时空细粒度动作识别模型,完成测试集的动作识别任务。 Official tags: 骨骼点识别. Sign-ups: 188.
Resources
What this competition gives you
Taken from the competition's official page.
- DataCompetition skeleton-point dataset stored in .npy format.
- TutorialOfficial PaddlePaddle baseline explanation video.
- TutorialPaddlePaddle high-level API tutorial video.
- Reference paperFSD-10 figure-skating dataset paper.
- Reference paperMotion-Centered Figure Skating Dataset paper.
- Reference paperTemporal convolutional networks action-analysis paper.
- Reference paperSpatial temporal graph convolutional networks paper.
- DataPaddleVideo video-model development kit and repository.
Preparation
From registration to a first submission
- 01
PaddlePaddle 2.0
- 02
骨骼点时序数据处理
- 03
动作识别模型
- 04
时空图卷积网络(ST-GCN)
- 05
NumPy .npy 文件读写
- 06
CSV 结果文件生成
Before you commit: 难点在于动作类别的差异往往只出现在少数几个帧中,但模型同时又必须保留完整时序信息来判断动作大类和处理“多义帧”。因此,不能只依赖单帧姿态特征,而要建模骨骼点随时间变化的细粒度差异。
Source
How this page was assembled
Competition information is structured from the official page. Scores are platform estimates for decision support; official rules take precedence.
- Official competition page
- Baidu AI Studio
- Last checked
- Not recorded