High-Fidelity Road Network Generation with Latent Diffusion Models

Authors: Jinming Wang, Hongkai Wen, Geyong Min, Man Luo

IJCAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on multiple datasets demonstrate the proposed Graph Walker can effectively generate high quality road networks from noisy and sparse trajectories, showcasing significant improvements over state-of-the-art. Extensive experiments conducted on diverse real-world datasets show that Graph Walker outperforms both appearance-based methods and state-of-the-art diffusion-based graph generation approaches, demonstrating significant improvements in generation accuracy and superior robustness with limited trajectory data and unusual road topologies present.
Researcher Affiliation Academia 1Department of Computer Science, University of Exeter 2Department of Computer Science, University of Warwick EMAIL, EMAIL
Pseudocode No The paper describes methods in prose and uses figures for architectural diagrams (e.g., Figure 2 and Figure 3) but does not include any explicitly labeled pseudocode or algorithm blocks.
Open Source Code Yes 1Code at: https://github.com/Jinming Wang/Graph Walker.
Open Datasets Yes Concretely, for an area within the city specified by a bounding box of certain size, we collect its road network data using Open Street Map [Open Street Map contributors, 2017], and aerial images from Google Maps [Google, nd].
Dataset Splits No The paper mentions using data from different cities and varying area sizes for training and testing, and varying the number of input trajectories, but it does not specify explicit training, validation, or test dataset splits (e.g., percentages or sample counts) for the synthesized trajectories.
Hardware Specification Yes measured on a single NVIDIA 4090 GPU
Software Dependencies No The paper does not provide specific version numbers for any software dependencies used for implementation (e.g., Python, PyTorch, TensorFlow, CUDA).
Experiment Setup No The paper describes the loss functions used for training the W2G-VAE and T2W-DiT, and mentions following 'standard training procedures for diffusion models', but it does not provide specific hyperparameter values such as learning rates, batch sizes, number of epochs, or optimizer details.