The paper proposes a novel segmentation strategy for kidney glomeruli detection. The algorithm was developed using a data set from the Kidney Pathology Image Segmentation (KPIs) challenge, which has the objective of accurate identification of glomeruli in chronic kidney disease (CKD) tissue samples. The proposed approach employs an ensemble of two SegNeXt models, building on the robust features of ConvNeXt. The ensemble approach effectively addresses the challenge related to variations in glomeruli size, shape, and structural integrity by leveraging hierarchical feature extraction. Integrating IoU loss, cross entropy loss, and entropy regularization improves segmentation performance by resolving class imbalance and promoting confident predictions. Additionally, extensive data augmentation strategies were used to enhance model robustness against variations in staining and image acquisition conditions. Experimental results demonstrate that the ensemble approach achieves high accuracy in segmenting glomeruli within localized image patches, showcasing its adaptability across diverse CKD scenarios. This work significantly contributes to pathological assessment of kidneys, providing a reliable tool for improving diagnostic accuracy and patient care.