Biliary tract localization using data augmentation techniques in endoscopic images for semi-autonomous robotic surgery applications

This thesis explores the application of data augmentation techniques, such as Generative Adversarial Networks (GANs) and various image transformations in order to improve the performance of the YOLO (You Only Look Once) algorithm in identifying and localizing the biliary tract in endoscopic images and to increase the diversity and robustness of the training data. The original and augumented datasets are used to train the YOLO model and metrics such as Intersection over Union (IoU) and mean Average Precision (mAP) are used to evaluate the effectiveness of the data augmentation techniques. The results show significant improvement in the model's ability to precisely localize the biliary tract, indicating that data augmentation can help alleviate the difficulties caused by incomplete and unbalanced medical datasets.