Social LSTM using PyTorch for Vehicle Data

This code/implementation is edited version of Anirudh Vemula's code. It is edited for vehicle trajectory data . If you are using this code for your work, please cite the original paper and Anirudh Vemula's original code.

Getting Started

The dataset available is normalized between -1 and 1. Also this version of code is only for GPU's.

Requirements

• Python 3.6
• Seaborn (https://seaborn.pydata.org/)
• PyTorch 0.4 (http://pytorch.org/)
• Numpy
• Matplotlib
• Scipy
• GPU

How to Run

• Before running the code, create the required directories by running the script make_directories.sh
• Unzip the data files inside the data_vehicles folder
• To train the model run python3 social_lstm/train.py (With default parameters)
• To test the model run python3 social_lstm/sample.py --epoch=n where n is the epoch at which you want to load the saved model. (Also since we use validation, by the end of training you should see the best epoch)
• To visualize and plot the grid run python3 social_lstm/visualize.py with default parameters