Toward Seasonal Guidelines for Robust Deep-Learning Sentinel-2 Building Detection in Different Area Types

2026-07-22

AI Summary

This study addresses the instability in small-scale building detection using Sentinel-2 imagery, which stems from its 10-meter spatial resolution and variations across seasons and built-up area types. The authors construct a multi-temporal Sentinel-2 dataset paired with official building vector labels and systematically evaluate the impact of seasonality, processing level (L1C vs. L2A), and regional heterogeneity on detection performance by fine-tuning U-Net and DeepLabV3+ models monthly and conducting cross-seasonal inference. For the first time, the work quantifies how seasonal dynamics and built-environment heterogeneity affect building extraction from Sentinel-2 data, identifies optimal months for training and inference, reveals cross-seasonal transferability patterns, and derives practical guidelines for model selection and context-specific operational deployment.

Abstract

Sentinel-2 imagery offers open access, global coverage, and frequent revisit times, making it attractive for practical building mapping at scale; however, its native 10m resolution makes building vs non-building classification challenging, particularly for small or sub-pixel buildings, and performance can vary with both seasonality and the heterogeneity of built-up environments. This paper introduces a Sentinel-2 building-detection framework designed to systematically quantify these effects and to support more formalised, practice-oriented model selection. We construct a dedicated multi-temporal Sentinel-2 dataset over the Warsaw region and derive binary ground-truth masks by rasterising official Polish topographic database (BDOT10k) building footprints onto the Sentinel-2 pixel grid. Using two established convolutional segmentation backbones (U-Net and DeepLabV3+), we first perform scene-specific fine-tuning to select a robust architecture and identify the best monthly models for L1C and L2A products separately. We then conduct cross-temporal inference by applying each best monthly model to all scenes, enabling an assessment of (i) which months provide favourable training and inference conditions, (ii) how performance transfers between seasons, (iii) the impact of processing level, and (iv) how these effects differ across built-up typologies. Based on these results, we provide practical guidance for routine Sentinel-2 building classification under varying acquisition periods and settlement characteristics.

Problem

• Sentinel-2
• building detection
• seasonality
• spatial resolution
• built-up heterogeneity

Innovation

• seasonal robustness
• Sentinel-2 building detection
• cross-temporal inference
• multi-temporal dataset
• built-up typology