cvxtz/ecg_heartbeat_classification

This repository is a set of standalone ECG heartbeat classification experiments, not a deployed service. The main workflow loads preprocessed MIT-BIH or PTB heartbeat CSVs, trains a 1D convolutional classifier, and reports test accuracy and F1; PTB transfer variants initialize from MIT-BIH weights, with one freezing convolutional layers. A separate EDA script summarizes and plots test heartbeats. Each experiment implements its own data preparation, model, and evaluation inline; the graph keeps those workflows together. Evidence is limited to the README and the sampled scripts.