Cancer is still one of the major causes of death worldwide, with blood cancer being especially deadly and destructive. Acute lymphoblastic leukemia (ALL) stands out from all other types of cancer, especially in children. Image classification is nowadays the biggest research field for detecting cancer. These papers work with four state-of-the-art pre-trained deep learning models: DenseNet201, InceptionV3, InceptionResNetV2, and VGG19, which are convolutional neural network (CNN) architectures. We hope to improve the accuracy of all diagnoses by combining powerful deep learning techniques with medical picture analysis. The research involved a dataset of 3256 peripheral blood smear images, which were preprocessed to improve quality and isolate WBCs. During this research, we found that the methods proposed in this study improve the models’ performance. We find the best performance in VGG19, with a validation accuracy of 98.66%.