Improving the detection efficiency of IRAND based on Convolutional Neural Network (CNN)

"CNN-Based Automated Event-Type Prediction and Test Dataset Generation for Separating Reactor Antineutrino and Cosmic-Muon Signals"
This repository contains the enhanced version of the IRAND-Sim-02 simulation package, developed for the automatic generation of test datasets and the classification of antineutrino and cosmic-muon events.
The original version (published previously (Mousavi, Mahdieh Sadat; Rahmani, Faezeh (2025), “IRAND-Sim-02”, Mendeley Data, V1, doi: 10.17632/vjtk7pycn2.1)) supported two source modes: **antineutrino-only** and **cosmic-muon-only**.
In this extended version, a **combined source mode** has been added, enabling mixed event generation with user-defined probabilities — a key innovation for automated benchmarking, deep learning evaluation, and detector studies.
A real-time connection between **Geant4** and **Python** is implemented through a TCP socket to enable online analysis, automated image generation, and (optionally) automatic event-type prediction using a trained CNN model.