Computer Vision News - March 2018

29 Women in Computer Vision Adriana, what is your work about? I only joined Facebook a few months ago in November 2017. So far I’ve been continuing the research that I did previously at the Montreal Institute for Learning Algorithms (MILA) as a postdoc. When I joined MILA, I started working on biomedical challenges from the deep learning perspective: how can we improve deep learning models to tackle the challenges posed by biomedical data? My most recent project was about graph structure data. How can we build deep learning models that would handle efficiently and would have a good performance dealing with the graph structure data. Why did you choose this subject? When I arrived in Montreal as a postdoc, it was a time when deep learning was starting to reach the biomedical domain. Although my background is in computer vision, I thought that we could make a great impact in our society. We could actually deal with biomedical problems, address them, and support doctors as well. My inspiration came from a collaboration with neuroscience researchers at McGill University. Instead of working with biomedical imaging volumes, they extract surface meshes of the brain, because they can implicitly coregister among different imaging modalities and different subjects. Dealing with this kind of data seemed challenging to me at the moment because data is not as regular as images or texts. It’s irregular data. It lies in the non-euclidean domain which poses different challenges, and I saw it as a new opportunity to learn things. It seems that was the result of a very adult dream. Is this a recent decision? My studies in the biomedical domain go a little bit further back. I actually started working on biomedical problems and biomedical challenges from an imaging perspective when I was a Master’s student. At that time, things were a bit different. We didn’t have access to Women in Science Computer Vision News Adriana Romero Adriana Romero is research scientist at Facebook AI Research and adjunct professor at McGill University as of very recently.

RkJQdWJsaXNoZXIy NTc3NzU=