Multi-Modal Dynamic Fusion with Training-Free Generalisation for Integrated Deepfake Detection

This repository contains the complete implementation of the proposed machine learning framework, including data preprocessing, model construction, training, testing, and evaluation procedures. The code is organized into different modules and branches to ensure a clear and reproducible workflow. It includes scripts for dataset preparation, model training, performance evaluation, and result generation, as well as the required model configurations and trained weights. The project is designed to support transparency and reproducibility, allowing researchers to replicate the experiments by installing the required dependencies, preparing the dataset, running the training process, and evaluating the generated results. The main subject areas of this work include machine learning, artificial intelligence, deep learning, computer vision, and data science. No external funding was received for this research. The provided code and documentation offer a complete reproducibility workflow for verifying the reported results and supporting further research development.本仓库收录了所提出的机器学习（Machine Learning）框架的完整实现代码，涵盖数据预处理、模型构建、训练、测试与评估全流程。代码按不同模块与分支进行架构划分，以保障工作流清晰且具备可复现性。项目包含数据集准备、模型训练、性能评估、结果生成相关脚本，以及所需的模型配置文件与训练所得的权重。本项目旨在提升研究透明度与可复现性，研究人员可通过安装必要依赖项、准备数据集、运行训练流程并评估生成结果，复现本项实验。本工作的核心研究领域包括机器学习、人工智能（Artificial Intelligence）、深度学习（Deep Learning）、计算机视觉（Computer Vision）与数据科学（Data Science）。本项研究未获得任何外部资助。本项目提供的代码与文档构建了完整的可复现工作流，可用于验证已报道的实验结果并支撑后续研究拓展。