透過您的圖書館登入
IP:18.226.222.12
  • 學位論文

用於人體姿勢估測之對抗式訓練方法

Self Adversarial Training for Human Pose Estimation

指導教授 : 陳煥宗

摘要


本篇論文提出一個基於深度學習,用於人體姿勢估測的方法。我們採用生成對抗網 路作為我們學習模式,其中我們建立了兩個堆疊式沙漏型網路,一個作為生成網 路,一個作為鑑別網路。訓練完成後,生成網路會直接當作人體姿勢估測使用。鑑 別網路用來區分標準答案的熱圖和生成的熱圖,以此計算出的對抗損失反向回饋至 生成網路。此過程能使生成網路學習人體姿勢合理性,且從實驗結果發現,這樣的 訓練有助於提升預測的準確度。

並列摘要


This thesis presents a deep learning based approach to the problem of human pose estimation. We employ generative adversarial networks as our learning paradigm in which we set up two stacked hourglass networks with the same architectures, one as the generator and the other as the discriminator. The generator is used as a human pose estimator after the training is done. The discriminator distinguishes groundtruth heatmaps from generated ones, and back-propagates the adversarial loss to the generator. This process enables the generator to learn the plausible human body congurations and is shown to be useful for improving the prediction accuracy.

參考文獻


estimation: New benchmark and state of the art analysis. In CVPR, 2014.
detection and articulated pose estimation. In CVPR, 2009.
abs/1701.07875, 2017.
[4] V. Belagiannis and A. Zisserman. Recurrent human pose estimation. CoRR,
abs/1605.02914, 2016.

延伸閱讀