AstroLoc: New SOTA Model for Space-to-Ground Image Localization
gabriberton · x · 2026-08-03
AstroLoc is a robust localization model for space-based imagery of Earth. It is trained jointly on astronaut photos from the ISS and satellite images.
- Core Innovation: It is the first APL model trained directly on 300k weakly labeled astronaut photos, solving the "needle in a haystack" problem of high-zoom localization.
- Performance: It sets a new SOTA on astronaut photography localization, satellite orbit determination (the "Lost in Space" problem), and digitized historical space shuttle imagery.
More from Research
- EMBL Launches AI Librarian: A Knowledge Layer for Life-Science Agents — EMBL-Rome · 2026-08-03
- NeurIPS Reviews a Lottery? Researcher Rants About Adversarial Reviewers and Unresponsive ACs — WhiteBear2018 · 2026-08-03
- NeurIPS 2026 to Host Workshop on Rigorous Foundations for LLM Interpretability — hugo_larochelle · 2026-08-03
- NeurIPS 2026 Workshop Call for Papers: Dynamic Alignment in Human-AI Coupled Systems — huashen218 · 2026-08-03
- The Blind Review Crisis in AI: arXiv Preprints Compromise Anonymity — deliprao · 2026-08-03
- Empirical Study: How AlphaFold is Actually Reshaping Science — soumitrashukla9 · 2026-08-03