Deep learning techniques have effectively treated about one million gastrointestinal patients in recent years. It is the most advanced medical imaging technique for the diagnosis of gastrointestinal illnesses such as ulcers, polyps, bleeding, and so on. Because manual diagnosis is time-consuming and difficult for medical practitioners, researchers have developed computational techniques for disease detection and classification. To overcome these issues, we present a capsule network variation that is less sophisticated but still robust and capable of extracting features for a better classification. Experimental results show that the proposed model can achieve 87.3%, 93.84% and 85.50% test accuracies on complex images such as CIFAR 10, fashion-MNIST and kvasir-dataset-v2 datasets, respectively. The performance of the proposed model is comparable to that of the state-of-the-art models on the datasets with a relatively small number of parameters.