search

HuynhLam / DAV

[ECCV 2020] Guiding Monocular Depth Estimation Using Depth Attention-Volume
Other
36 stars 3 forks source link

Question about ScanNet dataset #1

Open Tord-Zhang opened 4 years ago

Tord-Zhang commented 4 years ago

Hi,

Since the resolution of RGB frames and depth are different, is any alignment operation required? Or I just need resize both RGB image and depth to the same resolution?

Thanks.

HuynhLam commented 4 years ago

Hi,

TL;DR. Yes, the RGBs are bilinear downsampling to match the depth resolution.

We obtained the ScanNet data from the Robust Vision Challenge 2018. After running scripts to download, unzip, split and downsample the data, the resolutions of RGB and depth images are match (640x480). There is no alignment required in this case.