Fig. 4
From: Trustworthy deep learning for malaria diagnosis using explainable artificial intelligence
Workflow of the proposed malaria detection framework. The pipeline begins with input blood smear images, which undergo preprocessing steps including resizing, normalization, and augmentation, followed by stratified splitting into training and test subsets. In the training stage, two advanced hybrid CNN architectures Xception and Inception-ResNetV2 are fine-tuned using a two-phase strategy. The trained models are then evaluated on unseen test data, performing binary classification of red blood cells into Parasitized or Uninfected.
