ICCV Daily 2021 - Wednesday

year with Professor Pedro Sander. I realized that my first paper got accepted, and I was very excited to go to the conference to discuss this with people. But after that, I realized that my topic at that time was a little bit old-fashioned. The conference was elegant and very small, around 100 or 200 people. I hoped to come up with work with more impact. I asked myself what I should do. At the time, deep learning was something that many people were working on. It seemed that combining 3D with deep learning was quite an open program, so I decided to go in that direction. Isn’t it difficult to move so far away from home to follow your science? It’s not difficult for me because during high school I was in another city other than my home. It’s not difficult. [ laughs ] So you are used to it. What excites you so much about your work? As a visiting scholar at Professor Hao Su’s lab, doing the project with the multi-view stereo was a very exciting experience. The topic is very challenging. Given the multiple images, you should construct the underlying 3D shapes. The question is very fascinating. Just using the 2D input, you can pop out 3D. Getting the 3D, you can do a lot of things. You can rotate it and see how it actually looks like. The problem itself 23 DAILY ICCV Wednesday Songfang Han “ After I decided to do research in 3D deep learning, I researched the pioneer professors in this realm! ”