This repository contains the full implementation of the framework proposed in the paper: "Deep Learning-Based Detection of Bowel Sound Events in Continuous Recordings" The code includes model training, cross-validation, calibration, and post-processing steps used to reproduce the reported results. The implementation operates on precomputed HDF5 features. The raw audio preprocessing stage is described in the manuscript and is not included in this repository. The dataset used in this study is publicly available via Kaggle and is not redistributed here. Please refer to the original source for data access. This repository is provided to ensure reproducibility of the study.