README

Apollo cyber-record file parser (Python based)

Introduction

This tool presents multiple examples of how to parse data from a record file and save using predefined format. The tool has been built on example provided within Apollo (/apollo/cyber/python/examples/).

The samples provided here illustrate cases for, parsing data from:

• lidar: based on Velodyne VLS-128
• radar: based on Continental ARS-408-21
• camera: based on Leopard Imaging Inc's Camera - LI-USB30-AZ023WDRB

Files and functions:

The files/functions provided are as follows:

• record_parse_save.py: main function that parses record files from and saves extracted data to specified location
• parse_lidar.py: function to parse lidar data
• parse_radar.py: function to parse radar data
• parse_camera.py: function to parse camera data
• parser_params.yaml: YAML file with details of record-file location, where the output files should be saved and what sensor data should be parsed along with specific channel-names associated with particular sensor, it's configuration and it's location.
  • list of channels in a record-file can be obtained using cyber_recorder info. For example, to get details on a sample record file 20190422142705.record.00000 saved at: /apollo/data/record_files/2019-04-22-14-27-05/2019-04-22-14-27-05_records/ enter following at command prompt:
    • cyber_recorder info /apollo/data/record_files/2019-04-22-14-27-05/2019-04-22-14-27-05_records/20190422142705.record.00000
    • The output on screen will look like that presented in the image below:

Dependency

sudo pip install pyyaml

How-to-use:

• It is assumed that the user is within Apollo docker environment and has successfully built it. Please check documentation on Build Apollo if required.
• Modify parameters specified within parser_params.yaml to serve your purpose.
• After correct parameters are specified in the YAML file, run parser function in /apollo using:

./bazel-bin/modules/tools/record_parse_save/record_parse_save

• parsed sensor data will be saved in new folder along with associated capture or scan timestamps in a text file.

NOTES

• In the example setup here, parsed data is saved within the parent folder containing the records-file's folder. Every saved file has associated timestamp with it.
• All record-files within the folder are parsed. If any record file is corrupt, the parser will display error message and continue to next record file.
• radar data is saved in text files in JSON format for each scan
• lidar point-cloud data is saved in text files for each scan
• camera images are saved in jpeg file for each capture
• All timestamps are saved in a separate file with timestamp suffix, in the same order in which the parsed files are saved in corresponding folder.