Deep learning has been extremely ubiquitous in the field of computer vision. Deep learning has surpassed other approaches for image segmentation. Semantic image segmentation is the task of classifying each pixel in an image from a predefined set of classes. This paper introduces a simple model U-Net architecture that would filter the microscopic images. The proposed model includes copy connect (concatenation) layers, which is the backbone of ReLUNet architecture. The model possesses an encoding part (contraction part) and decoding part (expansive part) that capture the context with sampling layers and merging layers for a pixel-wise segmentation. The architecture is trained through a dataset of microscopic images and tested through their corresponding masks. Segmentation of a 256x256 image takes some micro seconds on a recent GPU. The U-net achieves 89.9% IoU (Intersection of Union) scores and 10% for testing the model in python IDE. It indicates that the proposed method has excellent performance in measurement accuracy and efficiency.