Large-scale management of surface water resources in urban areas can be difficult, especially if the region is subject to monsoonal waterlogging. Deep learning-based methods for computer vision tasks, such as image segmentation, can effectively be applied to remote sensing data for generating water body maps of large cities, aiding managerial entities, urban planners and policymakers. The robustness of these models to erroneous pixel level training class labels has not been studied in water body segmentation context. Label noise is commonly experienced in classification tasks and may hinder the performance. We collected and densely labeled Sentinel-2 images over Dhaka, one of the most densely populated and flood prone megacities in the world. We synthetically injected four types of label noise viz., i) Gaussian noise, ii) translation, ii) rotation, and iv) mirroring. Our primary objective is to observe and quantitatively analyze the effects of label noise on remote sensing data-driven deep learning models for water body segmentation. Our results show that salt and pepper noise (injected artificially using Gaussian noise) of only 50% can cause a massive 48.55% drop in Intersection-over-Union score. The consequences of learning from training data with different magnitudes and settings of label noise have been explored.

The accuracy of machine learning models in plant disease detection significantly relies on large volumes of knowledge-based labeled data; the acquisition of annotation remains a significant bottleneck in domain-specific research such as plant disease detection. While unsupervised learning alleviates the need for labeled data, its effectiveness is constrained by the intrinsic separability of feature clusters. These limitations underscore the need for approaches that enable supervised early disease detection without extensive annotation. To this end, we propose a self-supervised learning (SSL) framework for the early detection of soybean’s sudden death syndrome (SDS) using hyperspectral data acquired from an unmanned aerial vehicle (UAV). The methodology employs a novel distance-based spectral pairing technique that derives intermediate labels directly from the data. In addition, we introduce an adapted contrastive loss function designed to improve cluster separability and reinforce discriminative feature learning. The proposed approach yields an 11% accuracy gain over agglomerative hierarchical clustering and attains both classification accuracy and F1 score of 0.92, matching supervised baselines. Reflectance frequency analysis further demonstrates robustness to label noise, highlighting its suitability in label-scarce settings.