how would rotationnet deal with real 3d point cloud dataset?

[shangguan9191]

so, rotationnet use image. However, those images can not directly acquried from mesh data and point cloud data? then what is the point?

[kanezaki]

Images are rendered using MATLAB code. Please refer to: https://github.com/kanezaki/SHREC2017_track3/blob/master/mvcnn_utils/render_views.m

[shangguan9191]

If it is possible to give a example for transfer modlnet40 to image?

Asako Kanezaki [email protected] 于 2020年8月4日周二 上午8:57写道：

Images are rendered using MATLAB code. Please refer to:

https://github.com/kanezaki/SHREC2017_track3/blob/master/mvcnn_utils/render_views.m

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/kanezaki/rotationnet/issues/12#issuecomment-668417089, or unsubscribe https://github.com/notifications/unsubscribe-auth/APM2QBFPERJ2J6W2PWFAZCTR66WPFANCNFSM4PSPLAJA .

[kanezaki]

Please see "3. Render multi-view images" in https://github.com/kanezaki/SHREC2017_track3 This script is for the ShapeNet data, but it should work with ModelNet, too.