face point annotation
describe:
Existing face databases cover a wide variety, including: different subjects, poses, lighting, occlusions, etc. However, the annotations provided seem to have some limitations.
Figure 1: (a)-(d) Annotated images from MultiPIE, XM2VTS, AR, FRGC Ver.2 databases, and (e) an example of inaccurate annotations from XM2VTS.
- Most existing databases provide annotations for relatively small subsets of the entire image.
- In some cases, the accuracy of the provided annotations is not very good (probably due to human fatigue).
- The annotation models for each database consist of different numbers of landmarks.
These issues make cross-database experiments and comparisons between different methods almost infeasible. To overcome these difficulties, we propose a semi-automatic annotation method to annotate massive face datasets. This is the first attempt at creating a tool suitable for annotating massive face databases.
All annotations are for research purposes only (non-commercial product).
Figure 2: 68-point markers for annotation.
download:
We use our tool to create annotations for the following databases (following Multi-PIE 68-point notation, see Figure 2):
- 300-W [ Part 1 ][ Part 2 ][ Part 3 ][ Part 4 ]
Please note that the database is simply divided into 4 smaller parts for easier downloading. In order to create the database, you must unzip part 1 (ie 300w.zip.001) using a file archiver program (eg 7zip). - XM2VTS
- FRGC version 2
- low power pre-shock
- Helen
- AFW
- LOVE
reference:
Please cite as:
-
C. Sagonas, E. Antonikos, G. Tsimiropoulos, S. Zafiriou, M. Pantic. 300 Facing challenges in the wild: databases and results . Image and Visual Computing (IMAVIS), Special Issue on "In-The-Wild" Facial Landmark Localization. 2016.
-
C. Sagonas, G. Tsimiropoulos, S. Zafirio, M. Pantic. A Semi-Automatic Approach to Facial Landmark Annotation . Proceedings of IEEE International Conference. Computer Vision and Pattern Recognition (CVPR-W), 5th Symposium on Face and Gesture Analysis and Modeling (AMFG 2013). Oregon, USA, June 2013.
-
C. Sagonas, G. Tsimiropoulos, S. Zafirio, M. Pantic. 300 Faces in-the-Wild Challenge: The first facial landmark localization challenge . Proceedings of IEEE International Conference. Computer Vision (ICCV-W), 300-sheet challenge in the wild (300-W). Sydney, Australia, December 2013.
touch:
Christo Sagonas - [email protected] / Stefanos Zafeiriou - [email protected]