MARS - A foundational map auto-regressor
Quick Answer
This paper shows that MARS, the first map foundation model, generates multi-polyline road networks and polygon buildings using a unified auto-regressive approach.
Quick Take
Trained on the MAP-3M dataset with 3.4M examples, it outperforms multistage baselines across four datasets and features an interactive 'Chat with MARS' capability for map generation and correction.
Key Points
- MARS is capable of generating both road networks and building polygons.
- Trained on the largest multi-class map extraction dataset, MAP-3M, with 3.4M examples.
- Outperforms or matches multistage baselines across four datasets.
- Features an interactive 'Chat with MARS' for human-in-the-loop map generation.
- Dataset and project demo page are publicly available.
DeepSignal Analysis
What happened
MARS, a new map foundation model, generates multi-polyline road networks and polygon buildings using an auto-regressive approach. It was trained on the MAP-3M dataset, which contains 3.4 million examples, and shows improved performance over traditional multistage methods across four datasets.
Key evidence
- MARS is capable of generating both multi-polyline road networks and polygon buildings in a unified manner, addressing challenges in map generation tasks.
- The MAP-3M dataset, used for training MARS, is reported to be the largest multi-class map extraction dataset with 3.4 million examples.
- MARS outperforms or matches the performance of multistage baselines across four different road and building datasets.
Why it matters
The introduction of MARS could simplify the map generation process by reducing the complexity associated with traditional multistage methods. Its ability to generate detailed map features in a single step may lead to more efficient applications in urban planning, navigation systems, and geographic information systems. The interactive 'Chat with MARS' feature also suggests potential for user engagement in map correction and generation.
Paper Resources
📖 Reader Mode
~2 min readAuthors:Qi Zhang, Suvam Bag, Rupanjali Kukal, Lucas Thai, Jacob Kovarskiy, Mike Bianco, Xiaojie Guo, Raphael Tang, Mikael Figueroa, Rishi Madhok, Nikolaos Karianakis, Fuxun Yu
Abstract:Map generation tasks feature extensive non-structural vectorized data (e.g., points, polylines, and polygons) and thus pose significant challenges to common pixel-wise generative models. Conventional approaches use multiple stages, first segmenting these features at the pixel level and then performing vectorized post-processing, with errors and complexity compounding at each stage. Motivated by the recent success of auto-regressive language modeling, we propose the first map foundation model, named Map Auto-Regressor (MARS), that is capable of generating both multi-polyline road networks and polygon buildings in a unified manner. For training MARS, we collected to our knowledge the largest multi-class map extraction dataset totaling 3.4M examples, which we call MAP-3M. Across four road and building datasets, MARS outperforms or matches the performance of multistage baselines. Additionally, we develop a "Chat with MARS" capability that enables interactive human-in-the-loop map generation and correction, supported by the auto-regressive nature of our end-to-end approach. We release our MAP-3M dataset and project demo page at (1) this https URL and (2) this https URL, respectively.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2610.20843 [cs.CV] |
| (or arXiv:2610.20843v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.20843 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Rupanjali Kukal [view email]
[v1]
Wed, 14 Oct 2026 19:28:16 UTC (7,034 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →ProMoE-FL: Prototype-conditioned Mixture of Experts for Multimodal Federated Learning with Missing Modalities
ProMoE-FL introduces a Prototype-conditioned Mixture-of-Experts framework for multimodal federated learning, effectively addressing missing modalities. It outperforms existing methods on four chest X-ray datasets, demonstrating superior feature synthesis capabilities in both homogeneous and heterogeneous settings.