🚀 Projects

VR Museum

• For intelligent guidance in VR museum scenarios, this project integrates eye tracking, gaze estimation, and salient object detection to enable automatic artifact narration based on users’ visual attention.
• A multi-stage progressive training strategy was adopted to optimize Wav2Lip and DINet, improving lip-sync performance of digital humans; in addition, a Local-to-Global hierarchical salient object detection model based on spatial-channel attention was proposed, achieving an approximately 10% improvement in MAE over the SOTA.
• One software copyright has been filed.