Open WPC-WHU opened 5 months ago
Our current implementation requires 2D images, depth, and poses. If your dataset doesn’t include 2D images, you may need to render the 3D scene to obtain them. Once you have all this 2D data prepared, you can customize the functions get_mask_data
, get_seg_data
, and the code begining at line 234 in sai3d.py to suit your specific data.
If I only have 3D data (such as the S3DIS dataset), after I get the 2D images, how do I get the depth and poses?
How to segment my own dataset? 2D image is must?