Scholarlink

<p>Human pose estimation (HPE) is a fundamental challenge in computer vision, aiming to detect anatomical keypoints in images. Traditional methods rely on CNN models, but recent advancements in Vision Transformer (ViT) models have shown superior performance. However, ViTs often require substantial computational resources. This paper introduces SPTPose, a method that employs self-distillation and token pruning to reduce computational costs while maintaining high performance. Our SPTPose-B achieves a mAP of 74.8% on the MSCOCO validation set with only 13.2&#xa0;million parameters and 4.7 GFLOPs. The source code is available at <a href="https://github.com/duduxx123/SPTPose">https://github.com/duduxx123/SPTPose</a>.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhanced human pose estimation via self-distilled and token-pruned transformer

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月14日

查看全文
  • Jundu Zhang,
  • Zhengjie Deng,
  • Xiyan Li,
  • Sijian Yan,
  • Zhirui Li,
  • Chang Liu,
  • Weidong Ke

摘要

Human pose estimation (HPE) is a fundamental challenge in computer vision, aiming to detect anatomical keypoints in images. Traditional methods rely on CNN models, but recent advancements in Vision Transformer (ViT) models have shown superior performance. However, ViTs often require substantial computational resources. This paper introduces SPTPose, a method that employs self-distillation and token pruning to reduce computational costs while maintaining high performance. Our SPTPose-B achieves a mAP of 74.8% on the MSCOCO validation set with only 13.2 million parameters and 4.7 GFLOPs. The source code is available at https://github.com/duduxx123/SPTPose.