Dataset Viewer
The dataset could not be loaded because the splits use different data file formats, which is not supported. Read more about the splits configuration. Click for more details.
Couldn't infer the same data file format for all splits. Got {NamedSplit('train'): (None, {}), NamedSplit('validation'): ('json', {})}
Error code:   FileFormatMismatchBetweenSplitsError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

nuCarla 7-camera FOV30 extension

This repository adds a synchronized seventh camera to the six-camera nuCarla dataset. The new camera is CAM_FRONT_NARROW: a co-located 30-degree forward view derived from the original CAM_FRONT frame by a calibrated center crop and resize. This preserves the original scene, timestamp, objects, and six camera images.

Contents

  • 28,000 training samples and 6,000 validation samples.
  • 34,000 CAM_FRONT_NARROW JPEG images at 1600 x 900, distributed in 1,000-image TAR shards.
  • StreamPETR temporal info files containing all seven calibrated cameras.
  • Build lineage and validation reports.

Camera order:

CAM_FRONT
CAM_FRONT_NARROW
CAM_FRONT_RIGHT
CAM_FRONT_LEFT
CAM_BACK
CAM_BACK_LEFT
CAM_BACK_RIGHT

Required six-camera source

The six unchanged camera images are not duplicated in this extension. Extract the upstream nuCarla dataset so they are available below data/nucarla_full/samples, then place this repository at data/nucarla_full_7cam_fov30. The paths stored in the temporal info files will then resolve without rewriting.

After downloading this repository, extract its narrow-camera shards from the repository root:

for shard in image_shards/*.tar; do tar -xf "$shard"; done

This creates samples/CAM_FRONT_NARROW with the original 34,000 files and paths expected by the temporal info files.

The narrow-camera images are synchronized derivatives of the original frames; no independently rendered CARLA frame from another timestamp is inserted into the historical dataset.

Validation

validation_report.json verifies all 34,000 samples, camera ordering, intrinsics/extrinsics, 30-degree horizontal FOV, source hashes, image uniqueness, and transformed annotations.

The matching seven-camera StreamPETR checkpoint is published in shin0412/StreamPETR_nuCarla.

Downloads last month
39