First-Person Perspective Operational Video Dataset

Recording and analyzing information in the field of scientific research has always been a challenge. Current research video datasets on the market are mainly limited to third-person perspectives and lack first-person perspective videos that capture operator details and key points of operation, posing difficulties for recording and analyzing complex operations. To fill this gap, the First-Person Perspective Operational Video Dataset is focused on providing high-quality video material to improve action recognition and video understanding applications. The dataset was built using the latest shooting equipment, including head-mounted cameras and high-resolution video recorders, to collect multi-angle operational videos in a laboratory environment. Through multiple rounds of annotation, review by experienced researchers, and strict consistency checks, this dataset ensures high accuracy and high consistency. The professional annotation team is composed of experts in the field of information science and video processing, with a total number exceeding 50 people. During the data preprocessing, advanced video compression and enhancement technologies were adopted to ensure a balance between video quality and storage efficiency, and finally stored in MP4 format, organized by operation type and level of difficulty.