Histopathological images have significant potential for disease diagnosis and prognosis, but their inherent color variations can impede accurate analysis and reliable model performance. Color variations in histopathological images can indeed be a major challenge for accurate diagnosis and reliable model performance. This study investigates the critical role of color normalization in mitigating these inconsistencies, improving feature extraction, the model’s generalizability, and reducing staining bias. More this study examines the impact of activation functions, essential components in deep learning architectures for histopathological image analysis. Selecting the optimal function is crucial for optimizing the model performance. By emphasizing the importance of color normalization and careful activation function selection, this study lays the groundwork for more robust and reliable analysis of histopathological images, ultimately leading to improved disease diagnosis. In the present work, different segmentation architectures are analyzed with and without color normalization and with different activation functions.