Robotics Data Formats & Standards

RLDS (Reinforcement Learning Datasets) is Google DeepMind's standard format for robotics datasets. Understand its schema, which models use it, and how Claru delivers data in RLDS format.

HDF5 is the most widely used file format for robotics datasets. Learn its structure, compression options, and how Claru delivers robot training data in HDF5.

WebDataset uses tar archives for efficient sequential I/O in large-scale training. Understand the shard format, streaming capability, and Claru's WebDataset delivery.

TFDS provides structured dataset pipelines for TensorFlow. Learn how robotics datasets are structured in TFDS and how Claru delivers TFDS-compatible data.

LeRobot is Hugging Face's open-source robotics framework with its own dataset format. Learn the schema, ecosystem, and how Claru delivers LeRobot-compatible data.

Zarr provides chunked, compressed N-dimensional array storage ideal for large robotics datasets. Understand its structure and cloud-native capabilities.

ROS bags record timestamped sensor data from Robot Operating System. Learn the ROS 1/2 bag formats and how Claru converts between ROS and ML formats.

MCAP is a modern, high-performance container format for heterogeneous robotics data. Understand its advantages over ROS bags and how Claru uses MCAP.

The nuScenes format is the de facto standard for autonomous driving datasets. Learn its schema, annotation structure, and how Claru delivers compatible data.

The KITTI format is one of the most widely supported data formats in autonomous driving and 3D vision. Learn its file structure and how Claru delivers KITTI-compatible data.

BOP is the standard format for 6D object pose estimation benchmarks. Learn its structure and how Claru delivers pose estimation training data in BOP format.

COCO format is the most widely used annotation format for object detection and segmentation. Learn its structure and how Claru delivers robotics annotations in COCO format.

Open3D provides standard point cloud formats and processing tools for 3D robotics data. Understand PCD, PLY, and Open3D's tensor dataset format.

Apache Arrow and Parquet provide columnar data storage for efficient analytics and ML training. Learn how robotics tabular data is stored in Arrow format.

Protocol Buffers (protobuf) provide efficient binary serialization for robotics data schemas. Learn how protobuf is used in robotics pipelines and MCAP containers.