Reproduce “CARRADA Dataset: Camera and Automotive Radar with Range-Angle-Doppler Annotations” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

A. Ouaknine, A. Newson, J. Rebut, F. Tupin, P. Pérez

The CARRADA dataset provides raw range-angle and range-Doppler automotive radar data synchronized with camera recordings, accompanied by detailed range-angle-Doppler annotations for various road users. This resource enables the application of deep learning methods to raw radar signals, fostering research into robust, all-weather autonomous driving perception.

Key insights

  • Raw range-angle and range-Doppler radar data, when precisely annotated, can serve as effective inputs for deep learning models in autonomous driving perception tasks.
  • A semi-automatic annotation approach, combining information from complementary sensors like cameras with adaptive clustering on radar data, significantly reduces the effort for creating large-scale radar datasets.
  • The range-Doppler representation, which includes radial velocity information, provides more discriminative features for object classification and segmentation compared to the range-angle representation in radar data.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.