Introvert: Human Trajectory Prediction via conditional 3D attention

Overview

This repository contains the implementation of Introvert: Human Trajectory Prediction via conditional 3D attention.

In this work, we propose Introvert, a model which predicts human path based on his/her observed trajectory and the dynamic scene context, captured via a conditional 3D visual attention mechanism working on the input video.

Notes

The repository is still under construction.
Please let me know if you encounter any issues.

Best,
Nasim Shafiee

shafiee [dot] n [at] northeastern [dot] edu

---

Prerequisites

To install all the dependency packages, please run:

pip install -r requirements.txt

---

Data Preparation

We use UCY1 and ETH2 dataset. Please download and extract information into the ./data_trajpred folder. Click on here to download data

Now, in traj_pred.py find "# DataBase Variables" and update the data_traj_pred path.

---

Training and Testing

To run the code:

cd codes
CUDA_VISIBLE_DEVICES=0 python traj_pred.py --dataset="zara_01"
CUDA_VISIBLE_DEVICES=0 python traj_pred.py --dataset="zara_02"
CUDA_VISIBLE_DEVICES=0 python traj_pred.py --dataset="university"
CUDA_VISIBLE_DEVICES=0 python traj_pred.py --dataset="eth"
CUDA_VISIBLE_DEVICES=0 python traj_pred.py --dataset="hotel"

Citation

If you find the project helpful, we would appreciate if you cite the works:

@inproceedings{shafiee2021introvert,
  title={Introvert: Human trajectory prediction via conditional 3d attention},
  author={Shafiee, Nasim and Padir, Taskin and Elhamifar, Ehsan},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={16815--16825},
  year={2021}
}

---

References

[1] Lerner, Alon, Yiorgos Chrysanthou, and Dani Lischinski. "Crowds by example." Computer graphics forum. Vol. 26. No. 3. Oxford, UK: Blackwell Publishing Ltd, 2007. [2] Pellegrini, Stefano, et al. "You'll never walk alone: Modeling social behavior for multi-target tracking." 2009 IEEE 12th international conference on computer vision. IEEE, 2009.