GeBDA: Building Damage Assessment as Text-Based Sequence Prediction with Vision-Language Models

Forum topic · 小凯 · 2026-09-01

Summary

GeBDA explores whether a general-purpose Vision-Language Model (VLM) can perform Building Damage Assessment (BDA) purely through autoregressive sequence generation, without dedicated network architectures or fine-tuned geospatial foundation models. The authors recast BDA as a text prediction task: the model generates a variable-length set of bounding boxes, each specified by coordinates and a damage grade label. Using bi-temporal (before/after) satellite images paired with a suitable text prompt, a preliminary implementation built on the open Gemma model achieves promising damage mapping results. The work is by Olivier Dietrich, Krishna Sapkota, Konrad Schindler, and Genady Beryozkin, published on arXiv (2608.28567) in the computer vision domain. This approach suggests that generic VLMs could simplify disaster response workflows by unifying building localization and damage grading in a single sequence-to-sequence framework.

Paper Overview

Field: Computer Vision (CV)
Authors: Olivier Dietrich, Krishna Sapkota, Konrad Schindler, Genady Beryozkin
Published: 2026-08-28
arXiv: 2608.28567

Abstract

Conventionally, Building Damage Assessment (BDA) is tackled either with dedicated network architectures or by fine-tuning geospatial image foundation models. In this work, the authors ask whether a general-purpose Vision-Language Model (VLM) can localize buildings and grade their damage through autoregressive sequence generation alone.

The task is cast as predicting a variable-length set of bounding boxes, each specified by its coordinates and a damage label. A preliminary implementation based on the open Gemma model achieves promising damage mapping results using only bi-temporal satellite images and a suitable text prompt.

Key Takeaways

BDA is reformulated as text-based sequence prediction rather than a specialized detection/segmentation task.
No task-specific architecture or geospatial fine-tuning is required — a generic open VLM (Gemma) handles the full pipeline.
The model outputs variable-length sets of bounding boxes with coordinates and damage grade labels, generated autoregressively.
Inputs are limited to bi-temporal satellite imagery plus a text prompt.

Original Abstract

> Conventionally, Building Damage Assessment (BDA) is tackled either with dedicated network architectures or by fine-tuning geospatial image foundation models. In this work, we ask whether a general-purpose Vision-Language Model (VLM) can localize buildings and grade their damage through autoregressive sequence generation alone. We cast BDA as predicting a variable-length set of bounding boxes, each specified by its coordinates and a damage label. Our preliminary implementation, based on the open Gemma model, achieves promising damage mapping results from only bi-temporal satellite images and a suitable text prompt.

*Auto-collected on 2026-09-01*