Transfer Learning in Endoscopic Imaging: A Machine Vision Approach to GIT Disease Identification

In the context of rapidly rising Gastro-Intestinal tract related diseases among the population, the study explores the use of AI and ML that can enable early detection of the GI tract related abnormalities. The study after a detailed review of the existing studies, proposes on transfer learning to fine tune model weights trained on large data sets for GI-tract applications. The study has implemented models such as EfficientNet, MobileNet V3 Large, ResNet-50, DenseNet-201 and CNN (Convolutional Neural Network) on the endoscopic images. The top performing model, EfifcientNet2S, showcased superior proficiency identifying the eight distinct GIT diseases. The paper highlights the scope of applying transfer learning neural network architectures in enhancing disease diagnosis as well as selecting optimal architectures for accurate and efficient medical image classification in gastrointestinal pathology.