Malaria maintains its status as a serious public health issue which affects tropical and subtropical regions throughout the world. The condition causes the following symptoms: extreme tiredness and intense headache. The severe form of the disease can result in multiple organ failure or death. Timely and accurate diagnosis tropical and subtropical regions depend on timely diagnosis for effective treatment and control. The traditional diagnostic methods of blood smear microscopy and rapid diagnostic tests (RDTs) contain inherent limitations that arise from their use of subjective evaluation methods and their need for highly skilled technicians and their inconsistent detection capabilities. The study investigates advanced deep learning (DL) technology through VGG16 and ResNet50 models to perform automated malaria detection from blood smear images. The researchers used data augmentation and normalization techniques for both models under the same preprocessing conditions. This paper aims to conduct a direct comparison study that uses the same preprocessing techniques to evaluate the performance accuracy of the models. From the results, it is clear that the model, VGG16 (with transform), had the highest accuracy of 99.87 %, along with good recall and F1 scores. From the study, it is clear that ResNet50 (with transform) had an accuracy of 92.63 %, whereas ResNet50 (without transform) had an accuracy of only 58.55%. It is also clear that the ResNet50 (with transform) model had stable performance, whereas the ResNet50 (without transform) model had problems of overfitting and optimization. From the results, it is clear how the models' performance is enhanced using data augmentation techniques, as well as how deep learning technology can be used to detect malaria. It is clear how more datasets are required for the field, which will require more computing power as well as clinical understanding of the results. It is clear how more research needs to be done to optimize the models for better accuracy, especially for resource-constrained environments.