Computer Vision News - November 2019

3 Summary Hyo Jin Kim 31 Tower, I can tell you with my algorithm where it’s taken. This is by doing image retrieval against the geotagged images to estimate the position of the image using those references. At Facebook, I do similar work. It’s image-based localization as well as image retrieval, so we gather images that depict the same objects or buildings to assist 3D reconstruction. Facebook Reality Labs is focused on AR/VR. 3D reconstruction plays a big part. Our team focuses on something called LiveMaps. We are trying to create a virtual copy of the world. LG. The universities do joint projects with them focused on things like robot vacuums and other home electronics. Samsung are focusing more on AI, but before that they were focused more on SLAM and low-level visions like medical imaging. I think because of this explosion of deep learning, there is more of a focus on core machine learning right now. From the US, you get a different perspective. You are exposed to a wider variety of areas. That’s something that people in the US could bring back to Korea. Tell us about your work. During my PhD at UNC Chapel Hill I focusedonimageretrievalandmatching applied to image geo-localization. For example, if I have a picture of Namsan "We are trying to create a virtual copy of the world."