Complete blood cell count, which indicates the density of different blood cells in the human body is extremely important for evaluating the overall health of a person and also for detecting a wide range of disorders, including anemia, infection and leukemia. Hence, automating this task will not only increase the speed of diagnosis, but also lower the overall treatment cost. In this paper, we focus on using a convolution neural network to perform this complete blood cell count on blood smear images. The network is also trained to detect malarial pathogens in the blood, if present. Experiments show that the overall performance of the system has a mean average precision of over ${\bf 0.95}$ when compared with the ground-truth. Furthermore, the system predicts the images containing malarial parasites as infected ${\bf 100\%}$ of the time. The software is also ported to a low cost microcomputer for rapid prototyping.
