ABSTRACT
The gastrointestinal tract (GI tract) is a passageway in the body that allows nourishment to pass and solid waste to exit. The GI tract is comprised of several parts, including the throat, abdomen, small and large intestines, the anus, and the oesophagus. There are many different illnesses that can impact the gastrointestinal tract. These irregularities, if not identified early, have the chance of evolving into gastric cancer, a frequent condition of malignancy across over one million confirmed cases world wide each year. Endoscopy is an established technique for identifying and treating GI tract problems. Endoscopy is a microsurgical procedure used to examine the digestive tract. Gastroenterologists may miss some abnormalities during the examination for a variety of reasons, including irregular morphologies, a large number of frames, and exhaustion. Thus, the use of automated methods to categorize abnormalities found in endoscopic images is increasingly important in helping with medical diagnoses and cutting down on medical expenses and time. In this project. We intend to use the Kvasir medical datasets that are freely available to the public. It is made up of visuals from eight different gastrointestinal endoscopic imaging classes. Using Convolutional Neural Networks, we have simplified the recognition of GI tract markers and pathogens in Kvasir classes (CNNs). Because they can capture local features and are more highly scalable than fully connected networks, CNN models are broadly used in image perception.. We suggest using a dense model for the dataset that is based on the Densenet architecture. We strive to achieve a higher rate than that achieved in the reference papers.
