Biwi Kinect Head Pose Database with 15K Images

Filled with 15,000 images of head poses from 20 individuals, including 6 females and 14 males, with 4 subjects recorded twice. It was acquired using a Microsoft Kinect sensor, a structured IR light device, and is provided by ETH Zurich.

Use Cases

• Train a head pose estimation model using the 15,000 Kinect-acquired images of 20 subjects.
• Analyze pose variation across the 6 female and 14 male subjects to study demographic factors.
• Develop a person re-identification or verification system leveraging the data from 4 subjects recorded in two sessions.

Strengths

• Contains 15,000 images, providing a substantial volume for model training.
• Includes data from 20 distinct individuals, offering subject diversity.
• Annotations are expert-generated, suggesting high label quality.

Limitations

• Limited to 20 subjects, which may not capture the full diversity of head poses in a general population.
• The data was acquired with a specific sensor (Kinect), which may limit applicability to other camera types.
• No column or feature details are provided, making precise data structure unknown.

Provenance

Source

ETH Zurich

Collection Method

Acquired with the Microsoft Kinect sensor, a structured IR light device.

Freshness

Last updated on 2024-01-18.

Geography

Region tag indicates 'us'.