This paper introduces a noise augmentation technique designed to enhance the robustness of state-of-the-art (SOTA) deep learning models against degraded image quality. a common challenge in long-term recording systems. Our method. demonstrated through the classification of digital holographic images. https://www.duospiritalis.com/