Identifying the microbial population and type of them is a crucial measure in the water and wastewater treatment processes, reuse of wastewater, and sludge treatment system. Todays, manual methods are usually used to count and detect the type of bacteria in water and sewage laboratories which mostly suffer from human errors. This study aims at presenting an accurate method based on image analysis through the convolution neural network (CNN) to classify Escherichia coli (E. coli) and Vibrio cholera (V. cholera) bacteria, in wastewater. About 9,000 Red-Green-Blue (RGB) microscopic images of the sewage sample containing the stained bacteria were used as the input datasets. The results showed that the bacteria would be classified and counted with the accuracy of 93.01% and 97.0%, respectively. While CNN performed pretty well in counting the number of bacteria for both RGB and grayscale color models, its classification performance is only satisfactory in the RGB images. The sensitivity analysis of CNN illustrated that the Gaussian noise enhancement caused to the increment in the standard deviation () that proportionally decreased the CNN accuracy.