Scholarlink

Amid the growing prominence of Holographic Projection technology in Vietnam’s technology trend, we introduce NEURAHOLO: an innovative system designed to convert 2D images into 3D models for holographic displays. The system compares two advanced image processing algorithms for background removal U2-net and Mediapipe Selfie Segmentation followed by PIFuHD for 3D model generation, and a 3D Led Holodisplay for visualization. Our evaluation on the Supervisely Person Dataset with over 1000 images resulted in an IoU score of 0.9309 for U2-net, outperforming Mediapipe Selfie Segmentation, which scored 0.8467. The Chamfer Distance was used to measure the accuracy of the 3D models, with an average error of just 0.986, thereby corroborating the system’s adeptness at precise 3D model re-creation.

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

NEURAHOLO: A System for High-Resolution 3D Human Digitization on Holograms

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Hoang Pham Nguyen,
  • Thai Anh Huynh Ngoc,
  • Luat Le Gia,
  • Khoi Le Gia

摘要

Amid the growing prominence of Holographic Projection technology in Vietnam’s technology trend, we introduce NEURAHOLO: an innovative system designed to convert 2D images into 3D models for holographic displays. The system compares two advanced image processing algorithms for background removal U2-net and Mediapipe Selfie Segmentation followed by PIFuHD for 3D model generation, and a 3D Led Holodisplay for visualization. Our evaluation on the Supervisely Person Dataset with over 1000 images resulted in an IoU score of 0.9309 for U2-net, outperforming Mediapipe Selfie Segmentation, which scored 0.8467. The Chamfer Distance was used to measure the accuracy of the 3D models, with an average error of just 0.986, thereby corroborating the system’s adeptness at precise 3D model re-creation.