Open HardikJain02 opened 4 months ago
Our method requires point cloud, 2d image, depth and cam parameters as input. Also 3D fragment informantion like *.segs.json
in ScanNet is required for efficiency of the following region growing(Otherwise we have to implement per point region growing).
So if you only have point cloud on hand, you may need to render the required 2d data mentioned above, and generate 3D fragment data by yourself before running our algorithm.
You can refer to #6 to see how these data are loaded.
I am guessing a point cloud must be generated in the middle of the process. (as you need 2d image, depth, cam para)
Can I directly input point cloud?