This is a simple file reader for the Waymo Open Dataset modified from the simple-waymo-open-dataset-reader. The main goal is to start interacting with the dataset quickly without installing the waymo-open-dataset repo, which doesn't work for Mac. It does not aim to replace the Waymo repo, especially the evaluation metrics that they provide.
We additionally borrowed some preprocessing code from SceneInformer to convert motion dataset protobuf into a dictionary.
For dependencies, you only need protobuf, tensorflow:
pip install protobuf, tensorflow
Then install the repo:
pip install -e .
Optinally regenerate the *_pb2.py files from the protos:
sh generate_proto.sh
To parse a motion dataset example, download uncompressed/scenario/training/training.tfrecord-00000-of-01000 from thw Waymo website and then save it in a ./data folder. Then run:
cd examples
python parse_motion_dataset.py
You should see a new file parsed_data.p in the ./data folder.
Google could storage setup: You can download data directly from GCS without manually downloading the tfrecords. To set it up, first download the CLI package from here and go through items a, b, and c. Then follow the TF dataset authentication instructions here.
To run the above example using GCS:
cd examples
python parse_motion_dataset.py --gcs True
This code is released under the Apache License, version 2.0. This projects incorporate some parts of the Waymo Open Dataset code (the files simple_waymo_open_dataset_reader/*.proto) and is licensed to you under their original license terms. See LICENSE file for details.