CLiFF-LHMP is a pattern-based human motion prediction approach. Provided with maps of dynamics, the method predicts human motion in a long term.
- The following two datasets are supported:
- ATC pedestrian tracking dataset
- THÖR human motion trajectories dataset: containing THÖR1 and THÖR3 with different environment setup.
Part of both datasets are put in dataset
folder for demo. The dataset files provided are pre-processed. The downsample rate is 2.5 Hz. Maps are also included for visualizing the prediction results.
- Two prediction approaches are supported:
- CLiFF-LHMP: use maps of dynamics for prediction
- CVM: use constant velocity model for prediction
- To run the demo:
poetry run python main.py