comma-ai
Datasets
All datasets matching “comma-ai”commaCarSegments
commaCarSegments
commaCarSegments is a dataset of raw CAN bus data recorded from our fleet of openpilot users driving over 300 different production vehicles, around the USA and rest of the world.
Structure
.
├── segments/ # drectory containing all data
│ ├── <device_id>/ # unique device ID directory
│ │ └── <route_id>/ # unique route (aka drive directory)
│ │ └── <segment>/ # unique segment index inside of the route… See the full description on the dataset page: https://huggingface.co/datasets/commaai/commaCarSegments.comma2k19
comma2k19
comma.ai presents comma2k19, a dataset of over 33 hours of commute in California's 280 highway. This means 2019 segments, 1 minute long each, on a 20km section of highway driving between California's San Jose and San Francisco. comma2k19 is a fully reproducible and scalable dataset. The data was collected using comma EONs that has sensors similar to those of any modern smartphone including a road-facing camera, phone GPS, thermometers and 9-axis IMU. Additionally, the EON… See the full description on the dataset page: https://huggingface.co/datasets/commaai/comma2k19.comma1M
comma 1M
A large self-driving dataset containing one-minute driving segments with road-camera video, and full localization data.
Dataset collection
Segments were recorded by comma devices installed in real user vehicles. The collection spans several hardware generations (comma two, comma three, comma 3X, and comma four)
Camera systems vary between generations, so native resolution and field of view are not uniform across the dataset.
Each segment includes… See the full description on the dataset page: https://huggingface.co/datasets/commaai/comma1M.commaSteeringControl
commaSteeringControl
commaSteeringControl is a dataset of car steering measurements from ~12500 hours of driving with openpilot engaged. We control steering on most cars in openpilot using steeringTorque. This results in some lateral acceleration depending on both the car's internal vehicle dynamics and external factors (car speed, road roll, etc). Learning this relationship is essential to having accurate steering control in openpilot. commaSteeringControl is the largest controls… See the full description on the dataset page: https://huggingface.co/datasets/commaai/commaSteeringControl.commavq
commaVQ
commaVQ is a dataset of 100,000 heavily compressed driving videos for Machine Learning research. A heavily compressed driving video like this is useful to experiment with GPT-like video prediction models. This repo includes an encoder/decoder and an example of a video prediction model.
Examples and trained models can be found here: https://github.com/commaai/commavq
Overview
A VQ-VAE [1,2] was used to heavily compress each frame into 128 "tokens" of 10… See the full description on the dataset page: https://huggingface.co/datasets/commaai/commavq.commabody
comma body
A dataset of indoor navigation with the comma body.
description
This dataset is consists of 69 segments (1 min chunks) of video data compressed using H.265 and sensor measurements and logs from openpilot.
how to use
videos: use openpilot.tools.lib.framereader or your favorite video decoder
logs: use openpilot.tools.lib.logreader or PlotJuggler
have fun!