# Do Satellites See Commuters? A Critical Benchmark of Vision Foundation Models

arXiv: https://arxiv.org/abs/2609.00661 · HTML: https://mlsanity.com/paper/2609.00661/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 12 of its 22 references that are in the map.
- First posted on arXiv 1 Sep 2026.

## Builds on (most cited first, 12 of 12)

- [Denoising Diffusion Implicit Models](https://mlsanity.com/paper/2010.02502/) (2020)
- [Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks](https://mlsanity.com/paper/1703.03400/) (2017)
- [On First-Order Meta-Learning Algorithms](https://mlsanity.com/paper/1803.02999/) (2018)
- [RemoteCLIP: A Vision Language Foundation Model for Remote Sensing](https://mlsanity.com/paper/2306.11029/) (2023)
- [BigEarthNet: A Large-Scale Benchmark Archive For Remote Sensing Image Understanding](https://mlsanity.com/paper/1902.06148/) (2019)
- [SatCLIP: Global, General-Purpose Location Embeddings with Satellite Imagery](https://mlsanity.com/paper/2311.17179/) (2023)
- [AlphaEarth Foundations: An embedding field model for accurate and efficient global mapping from sparse label data](https://mlsanity.com/paper/2507.22291/) (2025)
- [Shift-Robust GNNs: Overcoming the Limitations of Localized Graph Training Data](https://mlsanity.com/paper/2108.01099/) (2021)
- [Deep Gravity: enhancing mobility flows generation with deep neural networks and geographic information](https://mlsanity.com/paper/2012.00489/) (2020)
- [Learning Geo-Contextual Embeddings for Commuting Flow Prediction](https://mlsanity.com/paper/2005.01690/) (2020)
- [Complexity-aware Large Scale Origin-Destination Network Generation via Diffusion Model](https://mlsanity.com/paper/2306.04873/) (2023)
- [Satellites Reveal Mobility: A Commuting Origin-destination Flow Generator for Global Cities](https://mlsanity.com/paper/2505.15870/) (2025)

## Abstract (arXiv)

Satellite foundation models offer a globally available alternative to census data for commuting origin-destination (OD) generation, yet no study has systematically compared encoder paradigms within a single downstream pipeline. We ablate four satellite vision encoders: language-supervised (RemoteCLIP), self-supervised (DINOv3), and geographically grounded (SatCLIP, AlphaEarth) within an identical WeDAN graph diffusion framework across 1,925 US counties, 325 UK districts, and 14 global cities under five random seeds. Three main findings emerge. First, language-supervised features achieve the strongest in-distribution performance (RemoteCLIP CPC 0.602), while geographically grounded encoders transfer more reliably zero-shot: AlphaEarth improves CPC by 33% over RemoteCLIP on UK districts. Second, pretraining corpus scale alone is insufficient: DINOv3, trained on a substantially larger satellite corpus, underperforms RemoteCLIP by 0.091 CPC in-distribution and collapses to CPC 0.022 globally. Third, no encoder transfers usefully to global cities (best CPC 0.122 for RemoteCLIP, 0.022 for DINOv3), confirming cross-continental OD generation remains an open problem. We additionally clarify the semantics of the census noise parameter η, whose ordering reverses under cross-continental evaluation, a distinction critical to correctly interpreting prior results. Training scripts and evaluation logs will be released.