metadata

dataset_info:
  features:
    - name: image
      dtype: image
    - name: depth
      dtype: image
    - name: label
      dtype: image
  splits:
    - name: train
      num_bytes: 152339736
      num_examples: 795
    - name: test
      num_bytes: 124128749
      num_examples: 654
  download_size: 276518209
  dataset_size: 276468485
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
task_categories:
  - image-segmentation

NYUDepth v2

Easier version for semantic segmentation.

Comes from:

Indoor Segmentation and Support Inference from RGBD Images (ECCV 2012) (PDF) (Bib)