Time-Series Anomaly Detection for Non-Invasive Fetal ECG using Diffusion Transformer

Overview

This repository provides a DiT-based NIFECG anomaly detection code developed as part of the ADVANCED GENERATIVE MODELS (STA8610) course project at Yonsei University Graduate School.

Features

• Anomaly NIFECG dataset and generation script
• 1D DiT-based base diffusion model
• DDAD-based reconstruction module

Installation

Clone this repository: git clone https://github.com/Minjee-Seo/AnomalyDiT.git

Usage

1. Edit the config.yaml file in ddad_utils.
2. Run python train.py for training.
3. Run python test.py for sampling and evaluation.

Contact

For any queries, please reach out to Minjee Seo and Seohyeon Jeong.