maturk
What's the proper way to use dn-splatter for processing data from my rgb+mask camera?
In my custom dataset, there are some moving people, so I generated corresponding masks for them. Typically, in COLMAP, I use the --ImageReader.mask_path parameter in feature_extractor to specify the masks.
How should these masks be incorporated into the dn-splatter workflow?
Hi, good question. Firstly, regarding your data, do you have colmap data? (i.e. a colmap/sparse/0 folder) and masks in some /masks folder? Or do you have a transforms.json file in nerfstudio format?
The code supports training with masks, it is just the matter of loading them in the dataloaders appropriately before training starts.
@maturk Yes, I have colmap data (colmap/sparse/0 folder) and masks in /masks folder. Not exist transfroms.json file. I want to know how to quickly and stably modify the code to import masks into the dataloader for filtering images. Do you have any reference scripts you can provide? Thank you.
For colmap data, you can use the --masks-path to give directory of masks that have the same name as images. You can use --help to see other options.
For example, you can try the command: ns-train dn-splatter coolermap --data [path to colmap data] --masks-path [path to masks folder].
Let me know if it works or not. You can also send me an example of your data (google drive) and I can see if it works or not. I have not used masks often with dn-splatter.