README.md

metadata

dataset_info:
  features:
    - name: image
      dtype: image
    - name: seg
      dtype: image
    - name: caption
      dtype: string
  splits:
    - name: train
      num_bytes: 590727018.85
      num_examples: 2975
  download_size: 583013818
  dataset_size: 590727018.85
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: apache-2.0
language:
  - en

Dataset Card for Dataset Name

Extract Cityscapes dataset (https://www.cityscapes-dataset.com/) training images and their semantic maps. Captions are generated with the pre-trained BLIP-large model.

Dataset Details

2975 training images and its semantic maps with auto-generated captions.