Three-Dimensional Trajectory Prediction with 3DMoTraj Dataset
Authors: Hao Zhou, Xu Yang, Mingyu Fan, Lu Qi, Xiangtai Li, Ming-Hsuan Yang, Fei Luo
ICML 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments show that our method significantly improves 3D trajectory prediction accuracy and outperforms state-of-the-art methods. Both the 3DMo Traj dataset and the method are available at https://github.com/zhouhao94/3DMo Traj. |
| Researcher Affiliation | Collaboration | 1School of Computing and Information Technology, Great Bay Institute for Advanced Study/Great Bay University, Dongguan, China. 2Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, China. 3Dongguan Key Laboratory for Intelligence and Information Technology, Dongguan, China. 4State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences, Beijing, China. 5Institute of Artificial Intelligence, Donghua University, Shanghai, China. 6Insta360 Research, Shenzhen, China. 7Nanyang Technological University, Singapore. 8University of California Merced, Merced, USA. |
| Pseudocode | No | The paper describes the proposed method, including decoupled trajectory prediction and correlated trajectory refinement, using prose and mathematical equations. However, it does not present these procedures in a structured pseudocode or algorithm block format. |
| Open Source Code | Yes | Both the 3DMo Traj dataset and the method are available at https://github.com/zhouhao94/3DMo Traj. |
| Open Datasets | Yes | To this end, we contribute a 3D motion trajectory (3DMo Traj) dataset collected from unmanned underwater vehicles (UUVs) operating in oceanic environments. ... Both the 3DMo Traj dataset and the method are available at https://github.com/zhouhao94/3DMo Traj. |
| Dataset Splits | Yes | The fragments for each scenario are then randomly divided into training, validation, and test sets with a 1:1:1 ratio. ... Furthermore, we recommend evaluating prediction methods using a leave-one-out cross-validation strategy. Specifically, the model is trained and validated on the training and validation sets of seven scenarios and tested on the test set of the remaining scenario. |
| Hardware Specification | Yes | All models in this experiment are tested on an NVIDIA 2080 Ti GPU using an input size of 70 8 3, where 70 represents the number of agents predicted simultaneously-exceeding the agent count in most real-world applications. |
| Software Dependencies | No | The paper does not explicitly mention specific software dependencies, such as programming languages or libraries, with their version numbers that are needed to replicate the experiment. |
| Experiment Setup | Yes | Training is conducted with a batch size of 70 for 100 epochs, employing the Adam optimizer with an initial learning rate of 0.0005. The learning rate decayed by 0.5 after the 5th epoch. Results are evaluated using (Average Displacement Error) ADE and (Final Displacement Error) FDE metrics. |