question about inference on custom real-world data

[LeiiY]

Thank you for your excellent work [Generalizable-Human-Gaussians]

I want to run the model inference on my custom real-world data, I wonder if the xxx.obj is necessary. in other words, in my real world data, I can obtain the following data:

1. RGB and depth imgs obtained by RGBD cameras.

2. intr and extr of RGBD cameras obtained by calibration

3. mask imgs obtained by matting algorithm

Is​ the data above enough to run the model inference like GPS-Gaussian? if so, how can I generate pos/vis maps without xxx.obj file?