Motion Forecasting in Continuous Driving

Authors: Nan Song, Bozhou Zhang, Xiatian Zhu, Li Zhang

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on Argoverse series with different settings demonstrate that our Real Motion achieves state-of-the-art performance, along with the advantage of efficient real-world inference.
Researcher Affiliation Academia Nan Song1 Bozhou Zhang1 Xiatian Zhu2 Li Zhang1 1School of Data Science, Fudan University 2University of Surrey
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks (clearly labeled algorithm sections or code-like formatted procedures).
Open Source Code Yes https://github.com/fudan-zvg/Real Motion
Open Datasets Yes We assess the performance of our method using the Argoverse 1 [3] and Argoverse 2 [43] motion forecasting datasets in both single-agent and multi-agent settings.
Dataset Splits Yes We conduct ablation studies on the Argoverse 2 validation split for the single-agent setting to examine the effectiveness of each component in Real Motion.
Hardware Specification Yes We measure these approaches and Real Motion on the Argoverse 2 test set using an NVIDIA Ge Force RTX 3090 GPU, maintaining a batch size of 1 and following an end-to-end manner.
Software Dependencies No The paper mentions using 'Adam W Optimizer' but does not provide specific version numbers for software dependencies or libraries like Python, PyTorch, or CUDA.
Experiment Setup Yes We train our models using the Adam W [24] Optimizer with a batch size of 32 per GPU for 60 epochs. Our model is trained end-to-end with a learning rate of 0.001 and a weight decay of 0.01. The latent feature dimension is set to 128.