Large-Scale Structure from Motion with Semantic Constraints of Aerial Images, PRCV 2018

Semantic 3D Reconstruction with Learning MVS and 2D Segmentation of Aerial Images, Applied Sciences 2021

RNN is trained to reconstruct the original vector representation of perceived words from the neural activity recorded by fNIRS.

Claude skill to proofread & semantically reconstruct error-dense medical ASR transcripts — multi-transcript diff, never fabricates, powered by a self-growing Notion glossary that learns confusing terms automatically. ｜ 醫學語音逐字稿校稿與語意重建的 Claude skill：多稿交叉比對、絕不捏造，搭配「會自動記錄易混淆字詞、自我成長」的 Notion 詞庫。