SLUM-i: Semi-supervised Learning for Urban Mapping of Informal Settlements and Data Quality Benchmarking

TL;DR

This work tackles the dual challenges of scarce annotations and semantic ambiguity in mapping informal urban settlements via remote sensing. It introduces a verified Lahore dataset with benchmarks for Karachi and Mumbai, and a cross-city evaluation framework across eight cities, complemented by a semi-supervised segmentation method that combines Class-Aware Adaptive Thresholding (CAAT) with a Prototype Bank to enforce semantic consistency. The approach demonstrates strong cross-domain transfer, achieving competitive or superior performance with only a fraction of labeled data, especially when paired with a DINOv2 backbone, while highlighting limitations in boundary precision and zero-shot generalization. The findings suggest that integrating calibrated pseudo-labeling with global semantic prototypes can substantially improve slum mapping, with potential for multi-modal data to further resolve visual ambiguities in urban informality.

Abstract

Rapid urban expansion has fueled the growth of informal settlements in major cities of low- and middle-income countries, with Lahore and Karachi in Pakistan and Mumbai in India serving as prominent examples. However, large-scale mapping of these settlements is severely constrained not only by the scarcity of annotations but by inherent data quality challenges, specifically high spectral ambiguity between formal and informal structures and significant annotation noise. We address this by introducing a benchmark dataset for Lahore, constructed from scratch, along with companion datasets for Karachi and Mumbai, which were derived from verified administrative boundaries, totaling 1,869 $\text{km}^2$ of area. To evaluate the global robustness of our framework, we extend our experiments to five additional established benchmarks, encompassing eight cities across three continents, and provide comprehensive data quality assessments of all datasets. We also propose a new semi-supervised segmentation framework designed to mitigate the class imbalance and feature degradation inherent in standard semi-supervised learning pipelines. Our method integrates a Class-Aware Adaptive Thresholding mechanism that dynamically adjusts confidence thresholds to prevent minority class suppression and a Prototype Bank System that enforces semantic consistency by anchoring predictions to historically learned high-fidelity feature representations. Extensive experiments across a total of eight cities spanning three continents demonstrate that our approach outperforms state-of-the-art semi-supervised baselines. Most notably, our method demonstrates superior domain transfer capability whereby a model trained on only 10% of source labels reaches a 0.461 mIoU on unseen geographies and outperforms the zero-shot generalization of fully supervised models.

Figures (8)

• Figure 1: Dataset characterization across boundary complexity, feature contrast, and signal-to-noise ratio dimensions.
• Figure 2: Visual comparison of feature contrast between Zia Colony, Lahore, a verified informal settlement ZiaColony (left), and Township, Lahore, a formal planned development (right). Images from Google Earth Pro GoogleEarthPro2025
• Figure 3: Annotation precision measured by boundary displacement relative to image gradients across all datasets. Narrow distributions indicate high consistency.
• Figure 4: Representativeness analysis across three cities (Lahore, Karachi, Mumbai). The t-SNE feature topologies, edge density, and entropy distributions show that the sampled subset accurately preserves the characteristics of the full dataset.
• Figure 5: Overview of the proposed semi-supervised framework. The pipeline consists of a shared encoder-decoder network processing labeled ($x^l$), weak ($x^w$), and strong ($x^{s1}, x^{s2}$) views.