LPFF: 19,590 Large-Pose Flickr Face Portraits

19,590 high-quality portrait images of faces captured across large pose variations. LPFF is a dataset introduced for ICCV 2023 by Yiqian Wu, Jing Zhang, Hongbo Fu, and Xiaogang Jin. The dataset is designed to support the creation of realistic facial images and 3D face shapes using generative networks.

Use Cases

• Training generative networks for 2D face image synthesis based on large-pose portrait images.
• Developing 3D face shape reconstruction models based on high-quality portrait data.
• Evaluating the performance of face generators across challenging pose variations.
• Benchmarking computer vision algorithms for portrait analysis and manipulation.

Strengths

• Contains 19,590 high-quality real portrait images.
• Specifically focuses on large pose variations, a challenging condition for face models.
• Associated with a peer-reviewed ICCV 2023 publication.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Freshness should be verified; last metadata update was 2023-11-04.

Provenance

Source

Flickr

Freshness

Last updated 2023-11-04 08:51:51