commaSteeringControl.py

import datasets
import pandas as pd
import pyarrow as pa

PLATFORMS = ["ACURA_RDX_2020", "AUDI_A3_3RD_GEN"]

_URLS = [f'https://huggingface.co/datasets/commaai/commaSteeringControl/resolve/main/data/{platform}.zip' for platform in PLATFORMS]

_DESCRIPTION = "Steering controls dataset from cars running openpilot"
N = 600


class CommaSteeringControl(datasets.ArrowBasedBuilder):

  def _info(self):
    return datasets.DatasetInfo(
      description=_DESCRIPTION,
      features=datasets.Features(
        {
          "t": datasets.Array2D(shape=(1, N), dtype='float'),
          "latActive": datasets.Array2D(shape=(1, N), dtype='bool'),
          "steeringPressed": datasets.Array2D(shape=(1, N), dtype='bool'),
          "vEgo": datasets.Array2D(shape=(1, N), dtype='float'),
          "aEgo": datasets.Array2D(shape=(1, N), dtype='float'),
          "steeringAngleDeg": datasets.Array2D(shape=(1, N), dtype='float'),
          "steer": datasets.Array2D(shape=(1, N), dtype='float'),
          "steerFiltered": datasets.Array2D(shape=(1, N), dtype='float'),
          "roll": datasets.Array2D(shape=(1, N), dtype='float'),
          "latAccelSteeringAngle": datasets.Array2D(shape=(1, N), dtype='float'),
          "latAccelDesired": datasets.Array2D(shape=(1, N), dtype='float'),
          "latAccelLocalizer": datasets.Array2D(shape=(1, N), dtype='float'),
          "epsFwVersion": datasets.Array2D(shape=(1, N), dtype='string'),
        }
      )
    )

  def _split_generators(self, dl_manager):
    downloaded_files = dl_manager.download(_URLS)
    return [
      datasets.SplitGenerator(
        name=PLATFORMS[i],
        gen_kwargs={
          "files": dl_manager.iter_archive(downloaded_files[i])
        }
      ) for i in range(len(downloaded_files))
    ]

  def _generate_tables(self, files):
    for path in files:
      df = pd.read_csv(path[1])
      pa_table = pa.Table.from_pandas(df)
      # print(pa_table)
      yield path[0], pa_table

  # def _get_examples_iterable_for_split(self, split_generator):
  #   for path in split_generator.gen_kwargs['files']:
  #     df = pandas.read_csv(path[1])
  #     yield path[0], df


# ds = CommaSteeringControl(use_auth_token=True)
# dp = ds.as_streaming_dataset(split="AUDI_A3_3RD_GEN")
# for p in dp:
#   print(p)
#   break