<p>Mumbai, a coastal megacity with a tropical climate and hilly topography, is one of the most populated cities in the world with a sizeable portion of its population living in slums. The continual expansion of slum areas places significant strain on the government’s capacity to provide essential services and also challenges efforts of urban planning. This study explores the novel application of Multi-channel Quantum Convolutional Neural Networks (MQCNNs) for slum classification using Sentinel-2 satellite imagery. While traditional machine learning models and deep CNNs have been employed for informal settlement detection, they often struggle with class imbalance, high intra-class variability, and limited training data. We propose a hybrid quantum–classical model based on Multi-Channel Quantum Image (MCQI) encoding and evaluate its performance against a structurally analogous classical CNN. To address the significant class imbalance in the dataset, we apply a stratified sampling strategy based on spectral intensity binning. Our results demonstrate that the MQCNN achieves superior performance, particularly on the minority (slum) class, with a test accuracy of 94.8% and an F1 score of 0.933, outperforming the classical CNN baseline. Ablation studies confirm the importance of sampling strategy and label purity thresholds. Comparable improvements have also been reported for the Jakarta dataset. These findings underscore the potential of quantum learning architectures in remote sensing tasks with limited or imbalanced data.</p>