Reflection Removal via Realistic Training Data Generation

Youxin Pang, Mengke Yuan, Qiang Fu, Dong Ming Yan

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Scopus citations


We present a valid polarization-based reflection contaminated image synthesis method, which can provide adequate, diverse and authentic training dataset. Meanwhile, we enhance the neural network by introducing the reflection information as guidance and utilizing adaptive convolution kernel size to fuse multi-scale information. We demonstrate that the proposed approach achieves convincing improvements over state of the arts.
Original languageEnglish (US)
Title of host publicationACM SIGGRAPH 2020 Posters
ISBN (Print)9781450379731
StatePublished - Aug 15 2020

Bibliographical note

KAUST Repository Item: Exported on 2020-10-15
Acknowledgements: This work was supported by the National Key R&D Program of China (2019YFB2204104 and 2018YFB2100602). (Portions of) the research in this paper used the 'SIR2' Dataset made available by the ROSE Lab at the Nanyang Technological University, Singapore.


Dive into the research topics of 'Reflection Removal via Realistic Training Data Generation'. Together they form a unique fingerprint.

Cite this