Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Commute in Data: The comma2k19 Dataset

Harald Schafer, Eder Santana, Andrew Haden, Riccardo Biasini

TL;DR

comma2k19 introduces a dense, consumer-sensor highway driving dataset (~33 hours) augmented with road-camera, IMU, CAN, and raw GNSS data, plus an open-source GNSS processor Laika that achieves ~40% positioning improvement. The work emphasizes tightly coupled GNSS/INS/vision fusion for global pose estimation on commodity hardware and provides global camera poses via Mesh3D. It validates Laika's benefits and offers rich data (including raw GNSS and ORB features) to advance localization and high-definition highway mapping in low-feature environments. The dataset and tools set a foundation for scalable, reproducible research in high-precision mapping using readily available sensors.

Abstract

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. The dataset was collected using comma EONs that have sensors similar to those of any modern smartphone including a road-facing camera, phone GPS, thermometers and a 9-axis IMU. Additionally, the EON captures raw GNSS measurements and all CAN data sent by the car with a comma grey panda. Laika, an open-source GNSS processing library, is also introduced here. Laika produces 40% more accurate positions than the GNSS module used to collect the raw data. This dataset includes pose (position + orientation) estimates in a global reference frame of the recording camera. These poses were computed with a tightly coupled INS/GNSS/Vision optimizer that relies on data processed by Laika. comma2k19 is ideal for development and validation of tightly coupled GNSS algorithms and mapping algorithms that work with commodity sensors.

A Commute in Data: The comma2k19 Dataset

TL;DR

comma2k19 introduces a dense, consumer-sensor highway driving dataset (~33 hours) augmented with road-camera, IMU, CAN, and raw GNSS data, plus an open-source GNSS processor Laika that achieves ~40% positioning improvement. The work emphasizes tightly coupled GNSS/INS/vision fusion for global pose estimation on commodity hardware and provides global camera poses via Mesh3D. It validates Laika's benefits and offers rich data (including raw GNSS and ORB features) to advance localization and high-definition highway mapping in low-feature environments. The dataset and tools set a foundation for scalable, reproducible research in high-precision mapping using readily available sensors.

Abstract

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. The dataset was collected using comma EONs that have sensors similar to those of any modern smartphone including a road-facing camera, phone GPS, thermometers and a 9-axis IMU. Additionally, the EON captures raw GNSS measurements and all CAN data sent by the car with a comma grey panda. Laika, an open-source GNSS processing library, is also introduced here. Laika produces 40% more accurate positions than the GNSS module used to collect the raw data. This dataset includes pose (position + orientation) estimates in a global reference frame of the recording camera. These poses were computed with a tightly coupled INS/GNSS/Vision optimizer that relies on data processed by Laika. comma2k19 is ideal for development and validation of tightly coupled GNSS algorithms and mapping algorithms that work with commodity sensors.

Paper Structure

This paper contains 12 sections, 1 equation, 5 figures, 1 table.

Table of Contents

  1. INTRODUCTION
  2. RELATED DATASETS
  3. Sample choice
  4. SENSOR SETUP
  5. Vehicles
  6. CAN messages
  7. Camera data
  8. Raw GNSS
  9. Other Sensors
  10. LAIKA
  11. GLOBAL POSES
  12. CONCLUSION

Figures (5)

  • Figure 1: Area driven in the comma2k19 dataset.
  • Figure 2: EON and grey panda installed in a vehicle.
  • Figure 3: Altitude error distributions for Laika and Live u-blox baseline algorithm for the two scenarios with antenna on the roof (left) and inside the car (right).
  • Figure 4: GNSS, INS and vision based 3D estimates of driven path and observed lanes projected onto video.
  • Figure 5: Viewing frusta from provided global poses (left) and global poses with map-based corrections (right).