Get a depth dataset from REPLICA dataset

[czeyveli1]

Hello everyone,

I am trying to create a metric depth dataset using the Replica RGB dataset, but I am not entirely sure how to proceed with it. I don’t quite understand which code file I need to work on to achieve this.

Additionally, below are the config settings for the dataset I will be using:

dataset_name: 'replica'
camera_params:
  image_height: 680
  image_width: 1200
  fx: 600.0
  fy: 600.0
  cx: 599.5
  cy: 339.5
  png_depth_scale: 6553.5
  crop_edge: 0

In which file exactly should I set these parameters?
In short, how can I obtain the depth outputs? I would appreciate it if you could help me with this.

[JUGGHM]

Well I think we need to configure two files:
(1) The json file to read the data, formatted like this. Here the camera intrinsics are configurated. The height and width do not need additional configuration.

(2) The metric_scale is configured here.

Before the training, we should check whether the depth is correctly loaded by adding some breakpoints here.