Stanford's Tempov Model Estimates Village Wealth from Satellite Imagery with 87% Accuracy
anselm · x · 2026-08-23
Stanford researchers released Tempov, a foundation model trained on three million pairs of Landsat images. It predicts asset wealth at the village level using raw satellite imagery (six spectral bands), updating in near real-time without census data or field surveys. The model explains 87% of household wealth variation in Malawi and 74% in Mozambique. Unlike existing models that predict static snapshots, Tempov focuses on tracking how wealth changes over time.
More from Research
- OpticalDNA: Rethinking Genomic Modeling via 2D Visual Representation — jiqizhixin · 2026-08-23
- Localsong: 1.2B Parameter Game Music Model Trained on One H100 — Amazing-You9339 · 2026-08-23
- New Checkpoint Released for Krea 2 Turbo 4-Step Distillation LoRA — TimeTruth2490 · 2026-08-23
- Experimenting with Nanbeige-Style Triple Loop on Qwen3.5-9B — Important-Farmer-846 · 2026-08-23
- Mathematics is About Much More Than Proof: Reflections on AI Automation — tobyordoxford · 2026-08-23
- MIT's new AI model SparksMatter discovers non-toxic, earth-abundant thermoelectric material — ProfBuehlerMIT · 2026-08-23