Why do frames within the same video have different image sizes and camera intrinsics??

[andrewsonga]

Hello Roman,

The CoP3D paper reported that each video of the CoP3D dataset consists of RGB frames of shape 3 x H x W, but the raw images provided as part of the actual dataset have different image sizes within the same video. Furthermore, the camera intrinsics (in units of pixels) differ between frames of the same video as well.

• Was there any preprocessing done on the raw video frames for the CoP3D dataset?
• Is the object scale the same across all frames of the released version of dataset?

It would be very convenient for myself (and many other people I assume) for the image sizes and intrinsics for frames within the same video to be constant. I was wondering if there was a way to achieve this with the CoP3D dataset.

Thank you :)

[Ir1d]

Hi @andrewsonga did you by chance figure this out already? Thanks!