MalNet – an Optimized CNN based method for Malaria Diagnosis
Swati Lipsa, Ranjan Kumar Dash · 2022 2nd International Conference on Intelligent Technologies (CONIT) · 2022
Due to the sheer advancement of modern computing, deep learning has emerged as one of the most promising technologies for machine learning. This work concentrates on applying deep learning to accurately diagnose disease. Our proposed method employs a Convolutional Neural Network (CNN) along with the optimal number and size of convolution and spooling layers. We have focused on a case study of the malaria diagnosis dataset containing pathological samples. This work utilizes Adam optimizer to train and validate the model. The choice of Adam optimization along with its important features is discussed. A suitable dataset is used and images are fed into the CNN without reducing their size or color and their performance is evaluated. An architectural comparison is performed between the proposed CNN model and some popular CNN architectures, resulting in a clearer vision of the proposed model's correctness while using a smaller number of hyperparameters. Furthermore, we also compared the proposed model to existing state-of-the-art models to assess its competency in terms of key performance metrics. This comparison reveals that this model's mechanism requires significantly fewer evaluation parameters, enabling the proposed method to become a time-efficient and computationally accurate model in terms of predicate accuracy.