aiqbench analysis
TrajGANR advances urban AI by unifying sparse GPS trajectory data with static street-view imagery through a continuous neural field, solving the alignment problem that plagues current geospatial models. This framework enables precise querying of mobility embeddings at arbitrary coordinates, significantly improving performance on fine-grained tasks like congestion prediction and road safety analysis in San Francisco and Porto. The approach demonstrates how integrating dynamic human navigation signals with visual context creates more robust representations for complex city-scale machine learning problems.
TrajGANR: Trajectory-Centric Urban Multimodal Learning via Geospatially Aligned Neural Representations
arXiv:2605.06990v2 Announce Type: replace-cross Abstract: Many urban prediction tasks depend not only on the static attributes of a location, such as the visual appearance of its built environment, but also on its dynamic function: how people navigate and use it over time. Predicting congestion, travel demand, or road safety risks, for instance, requires mobility information that imagery and geographic coordinates alone cannot provide. Trajectories carry this signal, yet current trajectory representations are difficult to align with static geospatial observations: sparse, irregular GPS samples rarely coincide with street-view image coordinates, and existing encoders produce either a single embedding per trajectory or road segment, or embeddings only at the observed GPS samples. We introduce TrajGANR, a trajectory-centric multimodal self-supervised learning framework that represents each trajectory as a continuous, path-conditioned neural field. Reconstructed from the observed GPS samples, this representation can be queried at arbitrary coordinates (including the off-path locations where street-view images were captured), yielding localized mobility embeddings that align with street-view image and location representations at the same coordinate. Across four fine-grained road-level tasks in San Francisco and Porto, TrajGANR consistently outperforms recent geospatial and trajectory foundation models, with ablations attributing the gains to both the mobility signal and the fine granularity at which it is aligned. TrajGANR's location encoder further generalizes beyond the road network to areas where neither street-view imagery nor trajectories are available.
Abstract:Many urban prediction tasks depend not only on the static attributes of a location, such as the visual appearance of its built environment, but also on its dynamic function: how people navigate and use it over time. Predicting congestion, travel demand, or road safety risks, for instance, requires mobility information that imagery and geographic coordinates alone cannot provide. Trajectories carry this signal, yet current trajectory representations are difficult to align with static geospatial observations: sparse, irregular GPS samples rarely coincide with street-view image coordinates, and existing encoders produce either a single embedding per trajectory or road segment, or embeddings only at the observed GPS samples. We introduce TrajGANR, a trajectory-centric multimodal self-supervised learning framework that represents each trajectory as a continuous, path-conditioned neural field. Reconstructed from the observed GPS samples, this representation can be queried at arbitrary coordinates (including the off-path locations where street-view images were captured), yielding localized mobility embeddings that align with street-view image and location representations at the same coordinate. Across four fine-grained road-level tasks in San Francisco and Porto, TrajGANR consistently outperforms recent geospatial and trajectory foundation models, with ablations attributing the gains to both the mobility signal and the fine granularity at which it is aligned. TrajGANR's location encoder further generalizes beyond the road network to areas where neither street-view imagery nor trajectories are available.
Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
arXiv:2605.06990 [cs.CV]
(or arXiv:2605.06990v2 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2605.06990 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 318 words
Related coverage
Where Will They Go? Modelling Multimodal Pedestrian Manoeuvres from Ego-centric Videos
arXiv · Jul 20, 2026
Context-aware Modality-Topology Co-Alignment for Multimodal Attributed Graphs
arXiv · Jun 15, 2026
DEXTERA: From a Single Image to Deployable Dexterous Manipulation via Real-to-Sim-to-Real
arXiv · Sep 21, 2026
Scaffolding Minds: Optimizing Latent Visual Target Representations for Multimodal Reasoning
arXiv · Aug 21, 2026