Question about Scannet Image Dataset

[pushkalkatara]

Hi, Thanks for the amazing work!

I wanted to ask for details about the Image dataset provided. For each scene there are following folders:

color/               depth-erode/         label-eroded/        labelv2/             pose/                
depth/               intrinsic_depth.txt  label-mapped/        labelv2-eroded/      world2grid.txt   

what are the differences between label-eroded, labelv2, and labelv2-eroded and depth, depth-eroded?

[Sekunde]

labelv2 is the scannetv2 annotations which are more complete than v1; eroded means the image has been applied by erosion operator so that the annotations are more smooth and less noisy