Computer Vision News - September 2019

8 Conclusion Vision scientists have long been interested in understanding how humans separate illumination and reflectance. Many previews works have explored this task. It is quite intuitive to try to use semantic information to tackle this problem and it seems that this research manages to do it very well. The semantic information extracted using state of the art deep learning techniques enables the method to produce high-quality IID results. It seems a promising method which will hopefully lead to a breakthrough in this task. Research