SharinGAN: Combining Synthetic and Real Data for Unsupervised Geometry Estimation

CVPR 2020  ·  Koutilya PNVR, Hao Zhou, David Jacobs ·

We propose a novel method for combining synthetic and real images when training networks to determine geometric information from a single image. We suggest a method for mapping both image types into a single, shared domain. This is connected to a primary network for end-to-end training. Ideally, this results in images from two domains that present shared information to the primary network. Our experiments demonstrate significant improvements over the state-of-the-art in two important domains, surface normal estimation of human faces and monocular depth estimation for outdoor scenes, both in an unsupervised setting.

PDF Abstract CVPR 2020 PDF CVPR 2020 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Monocular Depth Estimation KITTI Eigen split unsupervised SharinGAN absolute relative error 0.109 # 29
RMSE 3.77 # 2
Sq Rel 0.673 # 8
RMSE log 0.19 # 18
Delta < 1.25 0.864 # 18
Delta < 1.25^2 0.954 # 18
Delta < 1.25^3 0.981 # 17
Monocular Depth Estimation Make3D SharinGAN Abs Rel 0.377 # 2
Sq Rel 4.9 # 1
RMSE 8.388 # 5

Methods


No methods listed for this paper. Add relevant methods here