“…Deep learning-based approaches for counting are more precise and reproducible than conventional approaches because they can handle a wide range of objects with varying kinds, sizes, and complex materials and textures [ 29 ]. When we look at the literature on cell counting using automated tools, the bulk of the methods that are currently being used, however, rely on segmentation-based tactics, which call for a lot of training, tuning, and parameter optimization [ 30 , 31 , 32 , 33 ]. These techniques utilize image processing algorithms such as edge detection, thresholding, morphological operations, and watershed segmentation to separate cells from the background and from each other [ 34 , 35 , 36 , 37 , 38 ].…”