Hierarchical Motion Encoder-Decoder Network for Trajectory Forecasting (HMNet)
Our paper: https://arxiv.org/abs/2111.13324
We will release the complete version later
Setup:
The code was written in the following environment:
- python 3.7.11
- pytorch 1.10.0
- cuda 11.3
- cudnn 8.2.0
Preparation for data:
The raw data of Next Generation Simulation (NGSIM) is downloadable at https://ops.fhwa.dot.gov/trafficanalysistools/ngsim.htm
- Put the raw data into
./raw
- Run
preprocess_data.m
to preprocess the data for HMNet
Using the code:
To use the pretrained models at ./trained_models
and evaluate the models performance run:
python evaluate.py