Mirror Segmentation Dataset

Xin Yang,Haiyang Mei,Ke Xu,Xiaopeng Wei,Baocai Yin,Rynson W.H Lau.

Where is My Mirror?IEEE International Conference on Computer Vision (ICCV) 2019.

https://arxiv.org/pdf/1908.09101v2.pdf

Where is My Mirror.IEEE International Conference on Computer Vision(ICCV)2019..pdf

Please cite this paper when using the data： BibTex.txt

Introduction:

Mirrors are everywhere in our daily lives. Existing computer vision systems do not consider mirrors,  and hence may

get confused by the reflected content inside a mirror, resulting in a severe performance degradation. However,sepa-

rating the real content outside a mirror from the reflected content inside it is non-trivial. The key challenge lies in that

mirrors  typically reflect  contents  similar to their surroundings , making it very difficult  to differentiate the two. In this 

paper, we present a novel method to accurately segment mirrors from an input image.  To the best of our knowledge, 

this is the first work to address the mirror segmentation problem with a computational approach. We make the follow-

ing contributions.  First, we construct a large-scale mirror dataset  that contains mirror images with the corresponding 

manually  annotated  masks. This  dataset covers a variety of daily life scenes,  and will be made publicly vailable for 

future research.  Second , we propose a novel network , called MirrorNe t, for mirror segmentation , by modeling both 

semantical and low-level color/texture discontinuitiesbetween the contents inside and outside of the mirrors. Third, we 

conduct extensive experiments to evaluate the proposed method , and show that  it  outperforms the carefully chosen

baselines from the state-of-the-art detection and segmentation methods.