Malaria is an infectious sickness that influences a large number of lives each year. Customary conclusion of malaria in lab requires an accomplished individual and cautious investigation to segregate sound and contaminated red platelets (RBCs). It is likewise exceptionally tedious and may deliver wrong reports because of human mistakes. The target of this paper is to show how profound learning engineering, for example, convolutional neural organization (CNN) and Resnet-50 which can be valuable continuously malaria identification successfully and precisely from input pictures and to lessen difficult work with a portable application. To this end, we assess the presentation of a custom CNN model utilizing a repetitive stochastic inclination plummet (SGD) enhancer with a programmed learning rate locater and get an exactness in characterizing sound and tainted cell pictures with a serious level of accuracy and affectability. This result of the paper will work with microscopy determination of malaria to a portable application so unwavering quality of the therapy and absence of clinical mastery can be settled.