Closed lemonsstyle closed 2 weeks ago
Hi @lemonsstyle, I think the easiest would be to modify it in the format of ScanNet or you can write a wrapper that supplies the desired inputs to the model (which are basically camera poses, camera intrinsics, and RGB images)
Hi @lemonsstyle, I think the easiest would be to modify it in the format of ScanNet or you can write a wrapper that supplies the desired inputs to the model (which are basically camera poses, camera intrinsics, and RGB images)
Hello, thank you for your reply. I am familiar with the second method you mentioned, which is similar to MVS reconstruction datasets such as DTU and TNT (image folder, cams folder, camera parameters. txt)?
Right -- these inputs should work
Okay, I'll give it a try. Thank you very much!
Hello, thank you for your contribution. I would like to ask how to apply it to my own point cloud file for segmentation? Do I need to modify my point cloud file according to the format of a certain test dataset?