What question did this study set out to answer?

To resolve severe class imbalance and false-negative supervision caused by conventional sampling methods in remote sensing semantic segmentation by developing an object-centric patch sampling framework.

August 23, 2026Remote SensingOpen Access

Mitigating Class Imbalance and False-Negative Supervision in Remote Sensing Semantic Segmentation Using Object-Centric Patch Sampling

Q: What does this research mean for the field?

Object-centric patch sampling significantly improves semantic segmentation performance in remote sensing imagery by mitigating class imbalance and false-negative supervision compared to conventional sampling methods. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

View Full Paper

Ask AI

Bookmark

Authors

YRYogesh RegmiTribhuvan University SGSandeep GautamTribhuvan University GPGaurav ParajuliTribhuvan University

Discussion

Member takes

Implication

Experimental study demonstrates improved semantic segmentation across remote sensing imagery using object-centric patch sampling, highlighting a robust data-level fix for class imbalance.

Key Points

To resolve severe class imbalance and false-negative supervision caused by conventional sampling methods in remote sensing semantic segmentation by developing an object-centric patch sampling framework.
Designed a model-independent sampling strategy that anchors training patches to the geometric centroids of annotated objects and maintains a controlled proportion of background patches.
Evaluated performance across three datasets spanning Sentinel-2 satellite (cotton fields, 10 m), NAIP aerial (rural buildings, 1 m), and UAV imagery (water bodies, 0.25 m) using U-Net and DeepLabV3+ architectures.
Achieved IoU scores of 0.929, 0.896, and 0.912 on the three datasets using U-Net, outperforming sliding-window sampling by up to 19.6 percentage points in IoU.
Obtained a mean IoU of 0.9504 and a mean F1-score of 0.9772 in multi-class segmentation using DeepLabV3+, confirming architecture-agnostic improvements.

Cite This Study

Regmi et al. (2026) studied this question.

synapsesocial.com/papers/6a8aada77677a34114445f3a https://doi.org/10.3390/rs18162844

View Full Paper

Ask AI

Bookmark

Implication

Experimental study demonstrates improved semantic segmentation across remote sensing imagery using object-centric patch sampling, highlighting a robust data-level fix for class imbalance.

Mitigating Class Imbalance and False-Negative Supervision in Remote Sensing Semantic Segmentation Using Object-Centric Patch Sampling

Discussion

Member takes

Implication

Key Points

Cite This Study

Mitigating Class Imbalance and False-Negative Supervision in Remote Sensing Semantic Segmentation Using Object-Centric Patch Sampling

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider