HASTE: A No-Code Platform for Rapid Post-Disaster Building Damage Assessment from Satellite Imagery

Forum topic · 小凯 · 2026-07-15

Summary

HASTE (High-speed Assessment and Satellite Tracking for Emergencies) is a no-code web platform that lets analysts without machine learning expertise generate per-building damage maps from post-disaster satellite imagery within hours. Presented in arXiv paper 2607.11838 by Caleb Robinson, Anthony Ortiz, Simone Fobi Nsutezo, Cameron Birge, and Meygha Machado, the platform offers two methods sharing one interface: (1) users label polygons on a single post-disaster scene to train a small semantic segmentation model, whose per-pixel output is joined to existing building footprints; (2) a pretrained vision model embeds each building footprint, users label a few buildings, and a logistic regression model fitted in the browser scores the rest of the scene in seconds. Preliminary experiments on xBD show that foundation-model embeddings pooled over footprints can distinguish damaged from undamaged buildings using only post-disaster imagery, matching a fully supervised ResNet-50 baseline with roughly twenty times fewer labels. Since 2023, HASTE and its predecessors have supported over 30 real disaster responses—including earthquakes, hurricanes, cyclones, floods, wildfires, and tornadoes—delivering results to humanitarian partners within days of imagery availability.

Paper Overview

Field: Computer Vision
Authors: Caleb Robinson, Anthony Ortiz, Simone Fobi Nsutezo, Cameron Birge, Meygha Machado
Published: 2026-07-13
arXiv: 2607.11838

Problem

After a large disaster, responders need a map of damaged buildings within hours. Models that perform well on public benchmarks assume matched before-and-after imagery and training sets drawn from similar past events—neither of which is usually available on the first day of a new disaster.

HASTE Platform

HASTE (High-speed Assessment and Satellite Tracking for Emergencies) is a no-code web platform that enables analysts who are not machine learning engineers to produce per-building damage maps from post-disaster satellite imagery. It implements two methods that share a single interface:

1. Scene-level segmentation: The user labels polygons over the post-disaster scene; a small semantic segmentation model is trained on that single scene, run over the whole image, and the per-pixel output is joined to existing building footprints. 2. Embedding + lightweight classifier: A pretrained vision model embeds each building footprint; the user labels a small number of buildings, a logistic regression model is fitted in the browser, and the rest of the scene is scored within seconds.

Results

Preliminary experiments on xBD show that foundation-model embeddings pooled over building footprints can distinguish damaged from undamaged buildings using only post-disaster imagery, matching a fully supervised ResNet-50 baseline while requiring roughly twenty times fewer labels.

Real-World Deployment

Since 2023, HASTE and its predecessors have supported more than 30 real disaster responses across earthquakes, hurricanes, cyclones, floods, wildfires, and tornadoes, delivering results to humanitarian partners within hours to days after imagery becomes available.

Original Abstract (excerpt)

> When a large disaster strikes, responders need a map of which buildings are damaged within hours. The models that do well on public benchmarks assume matched before-and-after imagery and a training set drawn from similar past events, and neither is usually available for a new disaster in its first day. We present HASTE (High-speed Assessment and Satellite Tracking for Emergencies), a no-code web platform that lets analysts who are not machine learning engineers produce per-building damage maps from post-disaster satellite imagery. HASTE implements two methods that share one interface...

Full paper: https://arxiv.org/abs/2607.11838