This is a docker version of this, fixing some bugs and tutoring how to get the Waymo-motion-dataset, for quickly getting started with SceneTransformer, a trajectory prediction and planning network for autonomous vehicles.

ubuntu 20.04

cd /  && mkdir SceneTransformer_ws && cd SceneTransformer_ws
git clone [email protected]:bithuanglq/Docker-SceneTransformer.git

Create your image (e.g. scenetransformer:latest).

docker build -f Dockerfile -t scenetransformer .

You can create and run a container with the following command:

docker run -it -d --name container_name scenetransformer /bin/bash

And then, if you want to enter the container (to run commands inside the container interactively), you can use the docker exec command:

docker exec -it container_name /bin/bash

Put your validation dataset in /mnt/SceneTransformer/data/tf_records/validation/, the example validation data is uncompressed_tf_example_validation_validation_tfexample.tfrecord-00027-of-00150

cd /mnt/SceneTransformer 
python datautil/create_idx.py
python tmp.py

The Waymo moiton dataset is in form of structured dataset instead of images or BEV. Following this tutorial, installing the gcloud and gsutil command. Then get amounts of dataset using this web's gsutli command like below:

gsutil -m cp \
  "gs://waymo_open_dataset_motion_v_1_2_0/uncompressed/tf_example/validation/validation_tfexample.tfrecord-00000-of-00150" \
  "gs://waymo_open_dataset_motion_v_1_2_0/uncompressed/tf_example/validation/validation_tfexample.tfrecord-00001-of-00150" \
  .

Remember to put training dataset in dir: /mnt/SceneTransformer/data/tf_records/training/

Thanks to this repo, dataset, and paper.