README.md

metadata

dataset_info:
  features:
    - name: filename
      dtype: string
    - name: tfw
      dtype: binary
    - name: tif
      dtype: binary
    - name: label_tif
      dtype: binary
    - name: label_tfw
      dtype: binary
  splits:
    - name: train
      num_bytes: 2739248440
      num_examples: 180000
    - name: val
      num_bytes: 297399599
      num_examples: 19999
  download_size: 2804405571
  dataset_size: 3036648039
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: val
        path: data/val-*

Dataset Card for Sidewalks

This dataset comprises a collection of sidewalk images stored in a streamable format. It includes separate sets of Parquet files designated for training and validation, each with its respective split of data. Within the Parquet files, the dataset features information such as filenames, where image data is encoded as bytes, along with corresponding tfw files also encoded as bytes.

Code to stream train split

from datasets import load_dataset
dataset = load_dataset("back2classroom/sidewalks", split='test', streaming=True)
print(next(iter(dataset)))

Code to stream val split

from datasets import load_dataset
dataset = load_dataset("back2classroom/sidewalks", split='val', streaming=True)
print(next(iter(dataset)))