2023-01-012024-05-13https://scholars.lib.ntu.edu.tw/handle/123456789/646363子計畫二「多模態感知計算與互動實境」研究知覺計算與互動實境之知覺反饋機制，提供更自然且直覺的互動模式，促進元宇宙虛實融合相關技術之實現。在多模態感知計算方面，處理跨模態資料、無標註之大數據資料或小數據、可解釋性與隱私保護等挑戰。在互動實境方面，透過結合多種感知知覺資料，利用機器學習建立高階感知模型，並創造新工具帶入新的文化以及創意內容。主要研究關鍵議題為多模態智慧模型之自監督式學習與人本互動。 Subproject 2, "Multimodal Perception Computation and Interactive Reality," studies the perceptual feedback mechanism for perceptual computing and interactive reality, provides a more natural and intuitive interaction mode, and advances related technologies to integrate the virtual and real worlds. In multimodal perception computing, challenges include handling cross-modal data, unlabeled big or small data, explainability, and privacy protection. By integrating various perceptual data, machine learning can be used in interactive reality to establish advanced perception models and create new tools for introducing new cultures and creative content. Key research issues include self-supervised learning of multimodal intelligent models and human-centered interaction.