Computer Vision News - February 2020

3 Summary Revealing Scenes by Inverting ... 7 Another set of results from the paper is the importance of the RefineNet component. The next figure compares the output of CoarseNet with the output of RefineNet. It demonstrates the importance of using the latter and it shows how it dramatically improves the image appearance. "It performs quite well on the task. Well done!" The last result we show here is a novel view synthesis. This task is more popular in the computer vision community and it aims to generate an image from a new view, of the same scene. Generating a view synthesis with the method suggested in the paper is straightforward. Moreover, it can be seen below that it performs quite well on the task. Well done! The paper reveals a new issue of privacy in devices that perform localization. It does so by introducing a new technique to convert sparse 3D structure into realistic images of the scene. The method in the paper contains three networks to perform the task: one for visibility estimation, one for coarse image generation and one for refinement of the coarse image. The results of the paper are very good and demonstrate a concrete privacy issue in real world technologies! Conclusion "A concrete privacy issue in real world technologies!"