Optimizing Through Learned Errors for Accurate Sports Field Registration

Wei Jiang, Juan Camilo Gamboa Higuera, Baptiste Angles
Weiwei Sun, Mehrsan Javan, Kwang Moo Yi

Qualitative Highlights:

We apply our method to a broadcast video from YouTube. Besides reusing the optimization state for consecutive images, no temporal smoothing of any sort was applied. Although only trained with 170 training images, our method can generalize to games from different time, stadiums, and camera poses and is able to provide accurate registrations.

The supplementary videos are encoded by FFMPEG with h.264 codec.
If you can't play the video, please download the VLC player at: http://www.videolan.org/vlc/index.html