TY - GEN
T1 - CAFM
T2 - 2020 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops, WACVW 2020
AU - Sun, Yifan
AU - Murata, Noboru
N1 - Publisher Copyright:
© 2020 IEEE.
PY - 2020/3
Y1 - 2020/3
N2 - We present Cat-like Animals Facial Model (CAFM)- A 3D Morphable Model (3DMM) constructed from 50 samples, including lion, tiger, puma, American Shorthair, Abyssinian cat, etc. To the best of our knowledge, CAFM is the first animal morphable model ever constructed. New animal face images can be registered automatically by fitting pose and shape parameters of CAFM. Moreover, the parametric model regulates the naturalness of the generated animal faces avoiding unreasonable appearance.Computer vision has recently experienced great advances in automatic facial landmark detection. In this paper, to demonstrate CAFM's application to 3D reconstruction of cat face images, and to put effort towards uniform annotation scheme of immense databases and fair experimental comparison of cat-like animals' facial landmark systems, we improve the labeled cat face data set of10,000 images with 15 landmarks. Besides, we propose an algorithm matching our model to the input cat face images. With the projection parameters and shape parameter of CAFM, we can generate corresponding 3D meshes.
AB - We present Cat-like Animals Facial Model (CAFM)- A 3D Morphable Model (3DMM) constructed from 50 samples, including lion, tiger, puma, American Shorthair, Abyssinian cat, etc. To the best of our knowledge, CAFM is the first animal morphable model ever constructed. New animal face images can be registered automatically by fitting pose and shape parameters of CAFM. Moreover, the parametric model regulates the naturalness of the generated animal faces avoiding unreasonable appearance.Computer vision has recently experienced great advances in automatic facial landmark detection. In this paper, to demonstrate CAFM's application to 3D reconstruction of cat face images, and to put effort towards uniform annotation scheme of immense databases and fair experimental comparison of cat-like animals' facial landmark systems, we improve the labeled cat face data set of10,000 images with 15 landmarks. Besides, we propose an algorithm matching our model to the input cat face images. With the projection parameters and shape parameter of CAFM, we can generate corresponding 3D meshes.
UR - http://www.scopus.com/inward/record.url?scp=85085949350&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85085949350&partnerID=8YFLogxK
U2 - 10.1109/WACVW50321.2020.9096941
DO - 10.1109/WACVW50321.2020.9096941
M3 - Conference contribution
AN - SCOPUS:85085949350
T3 - Proceedings - 2020 IEEE Winter Conference on Applications of Computer Vision Workshops, WACVW 2020
SP - 20
EP - 24
BT - Proceedings - 2020 IEEE Winter Conference on Applications of Computer Vision Workshops, WACVW 2020
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 1 March 2020 through 5 March 2020
ER -