Scholarlink

The Scene2Model (S2M) tool, part of the OMiLAB Digital Innovation Environment, facilitates the digitalization of haptic storyboards created during Design Thinking workshops. However, the default scene element recognition system, based on QR identification, lacks flexibility, disrupting the creative process when new objects must be introduced. This work improves S2M by integrating a ResNet50Encoder-based One-Shot Learning (OSL) object recognition add-on. The proposed solution enables automatic recognition of previously unseen haptic objects through feature extraction and nearest-neighbor matching. By combining YOLOv4 for object localization and ResNet50Encoder for embedding extraction, this project improves storyboard capture flexibility while maintaining user control and flow during Design Thinking activities.

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

Enhancing the Scene2Model Tool: A ResNet50Encoder-Based Add-On for Flexible Object Recognition in Design Thinking

  • 章节

  • 发表日期：

    2026年

查看全文
  • Pierre Averty,
  • Dimitrios Kotzinos

摘要

The Scene2Model (S2M) tool, part of the OMiLAB Digital Innovation Environment, facilitates the digitalization of haptic storyboards created during Design Thinking workshops. However, the default scene element recognition system, based on QR identification, lacks flexibility, disrupting the creative process when new objects must be introduced. This work improves S2M by integrating a ResNet50Encoder-based One-Shot Learning (OSL) object recognition add-on. The proposed solution enables automatic recognition of previously unseen haptic objects through feature extraction and nearest-neighbor matching. By combining YOLOv4 for object localization and ResNet50Encoder for embedding extraction, this project improves storyboard capture flexibility while maintaining user control and flow during Design Thinking activities.