Activation Functions for Detection of Bacteria from Gram Stained Smears Images by YOLOv8

In this paper, we investigate the activation functions appropriate to detect bacteria from Gram stained smears images by YOLOv8. First, we detect bacteria by replacing all the occurrences of the default activation function SiLU at the convolution layers in YOLOv8 with other activation functions. Next, by focusing on the activation functions providing higher performance than SiLU to detect bacteria, we investigate which of occurrences of SiLU in several convolution layers in YOLOv8 replacing with the activation functions improves the results to detect bacteria.