Using kinect Xbox 360 for custom data

[MXNXV]

Hello Authors,
Thank you for the wonderful work, I want to try out your model on my custom data which was captured by a kinect Xbox 360 camera. My dataset looks as follows:
{ "files": [ { "rgb": "converted_data/250cm/rgb/r-1733262273.619000-4267990994.jpg", "depth": "converted_data/250cm/depth/d-1733262273.567000-4264455720.png", "cam_in": [ 586.34, 586.34, -6.77, -4.19 ], "depth_scale": 1000.0 },
I have captured depth maps and rgb images of a box at known distances. I am planning to use the pretrained weights to test out the model performance on my dataset.
The range of a kinect cameras depth sensor is from 1.2m to 4.5m. My question is should I use the default values in data_basic (mono/configs/HourglassDecoder/vit.raft5.small.py)
`data_basic=dict(

canonical_space = dict(

    # img_size=(540, 960),

    focal_length=586.34,

),

depth_range=(0, 1),

depth_normalize=(0.1, max_value),

crop_size = (616, 1064),  # %28 = 0

 clip_depth_range=(0.1, 200),

vit_size=(616,1064)

)`

[YvanYin]

You can use the default config. But you can adjust the 'clip_depth_range' to (0.1, 10). and 'depth_normalize' to (0.1, 10).