Detection & Classification of Abnormalities in GI Tract through MobileNetV3 Transfer Learning Model

The Gastrointestinal Tract is an essential organ system in the human body that plays a crucial role in food digestion, absorption, and elimination. Several Gastrointestinal Tract diseases or abnormalities may impact individuals' general health and well-being. Ulcers, polyps, and cancer are among the many illnesses and problems that can develop because of Gastrointestinal Tract disorders and abnormalities. Effective treatment and management of Gastrointestinal disorders depend on their early detection and precise classification. For this research, the MobileNetV3 Transfer Learning model with fine-tuning layers was implemented to identify and categorize Gastrointestinal Tract abnormalities. The model was trained and tested using the Kvasir dataset, which contains 3000 pictures of the Gastrointestinal Tract labeled into eight classes. The training, validation, and test sets were each allocated 8:1:1 of the total dataset respectively. The training set was used to fine-tune the MobileNetV3 model's performance on the given job during the training phase. The model's efficacy during training was tracked using the validation set, which helped avoid overfitting. The model's accuracy on the test set after training was 88%, demonstrating its accuracy in identifying and categorizing Gastrointestinal Tract abnormalities. This study illustrates the efficacy of employing MobileNetV3 and transfers learning to identify and categorize Gastrointestinal Tract problems. The model's excellent accuracy indicates that it has the potential to be a helpful tool for helping doctors identify and diagnose Gastrointestinal Tract disorders early on.