unitraj2hptr

Converts the h5 data from Unitraj (based on ScenarioNet) to the HPTR data format.

Installing

This tool can be installed by running the following commands:

conda create -n unitraj2hptr python=3.9.21
git clone https://github.com/marlon31415/unitraj2hptr.git
cd unitraj2hptr
pip install -e .

Preparing data

The conversion script expects .h5 Unitraj cache files as input. The original Unitraj repo converts the agent and map data to local coordinates. As HPTR expects global coordinates, we implemented a set of modifications necessary for the conversion pipeline which can be found in the modified Unitraj repo.

To preprocess the data follow this steps:

1. Install our modified ScenarioNet and convert data:

2. Install our modified Unitraj and generate cache files

3. Run the conversion script:

python convert.py --dataset split_to_convert --hptr_root_dir /path/to/hptr/output --unitraj_data_dir /path/to/input/data

Plotting

The conversion tool allows plotting scenarios while converting data, to activate this feature run:

python convert.py --dataset split_to_convert --hptr_root_dir /path/to/hptr/output --unitraj_data_dir /path/to/input/data --save_plots

NuScenes examples of default mode scenario HPTR outputs.