一些sota数据记录

  • Post author:
  • Post category:其他


数据来源:paperwithcode.com



三维人脸重建



数据集: Florence


Metric: Mean NME
model Mean NME



\downarrow











year
3DDFAv2 3.56% 2020
PRN 3.7551% 2018
VRN-Guided 5.2667% 2017
3DDFA 6.3833% 2015


Metric: Average 3D Error
model Average 3D Error



\downarrow











year
GANFit 0.95 2018
Unsupervised-3DMMR 1.50 2018
itwmm 1.82 2017
3DMM-CNN 1.93 2016


数据集:AFLW2000-3D


Metric: Mean NME
model Mean NME



\downarrow











year
PRN 3.9625% 2018
3DDFA 5.3695% 2015
DeFA 5.6454% 2017



人脸对齐



数据集: LS3D-W Balanced


Metric: AUC0.07
model AUC0.07 year
3D-FAN 72.3% 2017


数据集:AFLW


Metric:Mean NME
model Mean NME



\downarrow











year
3DDFAv2 4.43 2020
3DDFA 4.55 2018


数据集:AFLW2000-3D


Metric:Mean NME
model Mean NME



\downarrow











year
3DDFAv2 3.51% 2020
2DASL 3.53% 2019
PRN 3.62% 2018
3DDFA 3.79% 2018
3DSTN 4.49% 2017
DeFA 4.50% 2017
3DDFA+SDM 4.94% 2015


数据集:AFLW2000


Metric:Error Rate
model Error Rate



\downarrow











year
Nonlinear 3d face morphable model 4.70 2018
MCL 5.38 2018
3DDFA 8.250 2015
FAN 8.714 2017
Dlib 10.545 2014



版权声明:本文为thebestone1022原创文章,遵循 CC 4.0 BY-SA 版权协议,转载请附上原文出处链接和本声明。