Automatic Generation of 3D Videos from 2D Videos without Using Correspondence between Frames

2016 Fiscal Year Final Research Report

• Project/Area Number: 15H06669
• Research Category: Grant-in-Aid for Research Activity Start-up
• Allocation Type: Single-year Grants
• Research Field: Multimedia database
• Research Institution: Waseda University
• Principal Investigator: Iizuka Satoshi 早稲田大学, 理工学術院, 次席研究員(研究院講師) (30755153)
• Project Period (FY): 2015-08-28 – 2017-03-31
• Keywords: コンピュータグラフィクス / 3次元映像 / 画像処理 / ディープラーニング

Outline of Final Research Achievements

We have developed a fully convolutional neural network model that directly outputs a depth map from an image for automatically converting 2D videos to 3D videos. The technique can also be applied to videos by processing them frame-by-frame. Additionally, we have developed a method to easily generate complicated weathering effects in images and a technique to automatically colorize old black-and-white photographs with deep learning, which makes the generated 3D images and videos more photorealistic. These research results were accepted to the top conferences in the computer graphics field.

Free Research Field

コンピュータグラフィクス