Labeling Process
To address the lack of identity data, a custom-developed software was employed for manual labeling. Expert annotators identified and labeled faces, resulting in a rich dataset with over 3,000 distinct participants annotated from more than 40,000 faces across 8,837 images.
Key Features
- Source: pailixiang.com (Chinese photo live platform)
- Total Images: 8,837
- Total Faces: 40,000+
- Labeled Participants: 3,000+
- Annotations: Ground truth labels for face recognition
- Metadata: Upload time and number of views
Applications
This dataset is ideal for:
- Training and validating face recognition models
- Research in social event analytics and attendee identification
- Developing applications for real-time event monitoring and security
Â