Disentangled Multi-Relational Graph Convolutional Network for
Pedestrian Trajectory Prediction

Inhwan Bae · Hae-Gon Jeon
AAAI 2021

Project Page AAAI Paper Source Code Related Works

  

Left: Previous SOTA Model (CVPR'20)    Right: DMRGCN (Ours)

This repository contains the code for disentangling social interaction and alleviating accumulated errors for trajectory prediction.

🧶 DMRGCN Model 🧶

• Disentangled Multi-scale Aggregation for better social interaction representation on a weighted graph.
• Global Temporal Aggregation for alleviating accumulated errors when pedestrians change their directions.
• DropEdge technique to avoid the over-fitting issue by randomly removing relation edges.

Model Training

Setup

Environment
All models were trained and tested on Ubuntu 18.04 with Python 3.7 and PyTorch 1.6.0 with CUDA 10.1.

Dataset
Preprocessed ETH and UCY datasets are included in this repository, under ./dataset/. The train/validation/test splits are the same as those fond in Social-GAN.

Train DMRGCN

To train our DMRGCN on the ETH and UCY datasets at once, we provide a bash script train.sh for a simplified execution.

./scripts/train.sh

We provide additional arguments for experiments:

./scripts/train.sh <gpu_ids_for_five_scenes>

# Examples
./scripts/train.sh
./scripts/train.sh 0 0 0 0 0
./scripts/train.sh 0 1 2 3 4

If you want to train the model with custom hyper-parameters, use train.py instead of the script file.

python train.py --input_size <input_coordinate_dimension> --output_size <output_gaussian_dimension> \
--n_stgcn <number_of_gcn_layers> --n_tpcnn <number_of_cnn_layers> --kernel_size <kernel_size> \
--obs_seq_len <observation_sequence_length> --pred_seq_len <prediction_sequence_length> --dataset <dataset_name> \
--batch_size <minibatch_size> --num_epochs <number_of_epochs> --clip_grad <gradient_clipping> \
--lr <learning_rate> --lr_sh_rate <number_of_steps_to_drop_lr> --use_lrschd <use_lr_scheduler> \
--tag <experiment_tag> --visualize <visualize_trajectory>

Model Evaluation

Pretrained Models

We have included pretrained models in the ./checkpoints/ folder.

Evaluate DMRGCN

You can use test.py to evaluate our model.

python test.py --tag <experiment_tag>

# Examples
python test.py --tag social-dmrgcn-eth-experiment_tp4_de80
python test.py --tag social-dmrgcn-hotel-experiment_tp4_de80
python test.py --tag social-dmrgcn-univ-experiment_tp4_de80
python test.py --tag social-dmrgcn-zara1-experiment_tp4_de80
python test.py --tag social-dmrgcn-zara2-experiment_tp4_de80

📖 Citation

If you find this code useful for your research, please cite our trajectory prediction papers :)

💬 LMTrajectory (CVPR'24) 🗨️ | 1️⃣ SingularTrajectory (CVPR'24) 1️⃣ | 🌌 EigenTrajectory (ICCV'23) 🌌 | 🚩 Graph‑TERN (AAAI'23) 🚩 | 🧑‍🤝‍🧑 GP‑Graph (ECCV'22) 🧑‍🤝‍🧑 | 🎲 NPSN (CVPR'22) 🎲 | 🧶 DMRGCN (AAAI'21) 🧶

@article{bae2021dmrgcn,
  title={Disentangled Multi-Relational Graph Convolutional Network for Pedestrian Trajectory Prediction},
  author={Bae, Inhwan and Jeon, Hae-Gon},
  journal={Proceedings of the AAAI Conference on Artificial Intelligence},
  year={2021}
}

More Information (Click to expand)

@inproceedings{bae2024lmtrajectory,
  title={Can Language Beat Numerical Regression? Language-Based Multimodal Trajectory Prediction},
  author={Bae, Inhwan and Lee, Junoh and Jeon, Hae-Gon},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year={2024}
}

@inproceedings{bae2024singulartrajectory,
  title={SingularTrajectory: Universal Trajectory Predictor Using Diffusion Model},
  author={Bae, Inhwan and Park, Young-Jae and Jeon, Hae-Gon},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year={2024}
}

@inproceedings{bae2023eigentrajectory,
  title={EigenTrajectory: Low-Rank Descriptors for Multi-Modal Trajectory Forecasting},
  author={Bae, Inhwan and Oh, Jean and Jeon, Hae-Gon},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  year={2023}
}

@article{bae2023graphtern,
  title={A Set of Control Points Conditioned Pedestrian Trajectory Prediction},
  author={Bae, Inhwan and Jeon, Hae-Gon},
  journal={Proceedings of the AAAI Conference on Artificial Intelligence},
  year={2023}
}

@inproceedings{bae2022gpgraph,
  title={Learning Pedestrian Group Representations for Multi-modal Trajectory Prediction},
  author={Bae, Inhwan and Park, Jin-Hwi and Jeon, Hae-Gon},
  booktitle={Proceedings of the European Conference on Computer Vision},
  year={2022}
}

@inproceedings{bae2022npsn,
  title={Non-Probability Sampling Network for Stochastic Human Trajectory Prediction},
  author={Bae, Inhwan and Park, Jin-Hwi and Jeon, Hae-Gon},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year={2022}
}

Acknowledgement

Part of our code is borrowed from Social-STGCNN. We thank the authors for releasing their code and models.