Home Abstract Demo Paper Code


Depth estimation from a monocular 360˚ image is an emerging problem that gains popularity due to the availability of consumer-level 360˚ cameras and the complete surrounding sensing capability. While the standard of 360˚ imaging is under rapid development, we propose to predict the depth map of a monocular 360˚ image by mimicking both peripheral and foveal vision of the human eye. To this end, we adopt a two-branch neural network leveraging two common projections: equirectangular and cubemap projections. In particular, equirectangular projection incorporates a complete field-of-view but introduces distortion, whereas cubemap projection avoids distortion but introduces discontinuity at the boundary of the cube. Thus we propose a bi-projection fusion scheme along with learnable masks to balance the feature map from the two projections. Moreover, for the cubemap projection, we propose a spherical padding procedure which mitigates discontinuity at the boundary of each face. We apply our method to four panorama datasets and show favorable results against the existing state-of-the-art methods.


CVPR 2020

BiFuse: Monocular 360˚ Depth Estimation via Bi-Projection Fusion

Fu-En Wang*, Yu-Hsuan Yeh*, Min Sun, Wei-Chen Chiu, Yi-Hsuan Tsai

Paper Poster Code
    author = {Wang, Fu-En and Yeh, Yu-Hsuan and Sun, Min and Chiu, Wei-Chen and Tsai, Yi-Hsuan},
    title = {BiFuse: Monocular 360 Depth Estimation via Bi-Projection Fusion},
    booktitle = {The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month = {June},
    year = {2020}