djamelbenr/kalmanbaseline

Baseline Kalman models for trajectory predictions. One is constant velocity LSTM

Baseline Models

This is a code for baseline production using Kalman filter. It is the implementation of the models presented in : https://arxiv.org/abs/1908.11472

To use this code :

• Set the parameters, dataset path settings.yaml (The Bicycle model may show training unstabilities, contributions are welcome.)
• Run train_kalman_predict.py starts the trainning.
• Enter the name of the trained model in the load_name field of settings.yaml (should be in the form <model>_<dataset>_<id>)
• Run bokeh serve --show plot_interactive.py to open an interactive plot in the browser
• Run python save_results.py to save the prediction results computed on the test set
• Run python stats_results.py to print metrics evaluation, plot covariance matching and error histogram (from the saved results)

Datasets

NGSIM

From NGSIM website:

• Register at https://data.transportation.gov/Automobiles/Next-Generation-Simulation-NGSIM-Vehicle-Trajector/8ect-6jqj
• Download US-101-LosAngeles-CA.zip and I-80-Emeryville-CA.zip
• Unzip and extract vehicle-trajectory-data into raw/us-101 and raw/i-80

From googledrive:

• Download i-80: https://drive.google.com/open?id=19ovxiJLCnS1ar1dYvUIrfdz_UnGejK-k
• Download us-101: https://drive.google.com/open?id=14dMKew22_5evfOoSGEBYqHP3L92e7nyJ

Dataset fields:

• doc/trajectory-data-dictionary.htm

This Dataset is to be pre-processed with the Matlab function preprocess_data.m (that is a slightly modified version of the one from https://github.com/nachiket92/conv-social-pooling)

• Already preprocessed samples:
  • https://drive.google.com/open?id=171YZvK2DnJbtIgve6qAOkk6AFNrGAYTJ
  • https://drive.google.com/open?id=1N080iiQc43MTvLMqaLtsZh49okJL0hPx

Argoverse

Get the data from the Argoverse website:

• https://www.argoverse.org/

Put their API from Argoverse in a directory at same depth named "Argoverse"

• https://github.com/argoai/argoverse-api

Fusion

The Fusion dataset option relates to a private dataset that is not available.