همایش بین المللی هوش مصنوعی

صفحه اصلی / دومین همایش بین المللی هوش مصنوعی

Hybrid CNN-Based Deepfake Image Detection with Enhanced Cross-Forgery Generalization

نویسندگان :

Alireza Maleki¹ Hanie Zomorrodi² Hassan Khotanlou³ Atefeh Mosafer⁴

1- دانشگاه بوعلی سینا 2- دانشگاه بوعلی سینا 3- دانشگاه بوعلی سینا 4- دانشگاه بوعلی سینا

کلمات کلیدی :

Deepfake Detection،Feature Interaction Learning،Cross-Forgery Generalization،Hybrid Feature Extraction

چکیده :

In recent years, the rapid advancement of deep learning-based technologies for generating fake content has heightened the need for precise, reliable methods to detect fake images. In this study, an efficient method for distinguishing fake images from real ones using video frames is proposed. During preprocessing, video frames are isolated using a face-detection algorithm to improve the model's focus on the facial region. In order to extract complementary features at different levels of detail from the images, a parallel dual-branch architecture comprising the ResNet-101 and EfficientNet-V2 small models is utilized. The final classification between the real and fake image classes is achieved by concatenating the outputs of these two networks and feeding them into a four-layer Multi-Layer Perceptron (MLP). Two modules—Inter-Class Data Interaction and Dual-Domain Feature Interaction—are integrated to improve the model's generalization and accuracy in detecting various types of forgery. Results from experiments on a large dataset show that the proposed model performs admirably, achieving 98.26% accuracy and an AUC of 99.84%. Additionally, an analysis of the model's performance across various datasets shows that the suggested approach outperforms earlier approaches and exhibits strong generalization.

لیست مقالات

لیست مقالات بایگانی شده

An interpretable framework based on deep learning and the Internet of Things for predicting risk dynamics in chronic patients

Behnaz Pouriayevali - َAsghar Ehteshami

Towards Automated Self-Aware System Design: Process-Based Models to MAPE-K Architecture

Maryam Samsam Sokhiravi - Seyed-Hosein Attarzadeh-Niaki

Data Mining's Role in Crafting Intelligent Recommender Systems: A Systematic Review

Pourya Rahat - Amir Reza Asnafi

Interpretable Machine Learning for Rocking-Induced Settlement Prediction Using SHAP Analysis

Seyed Emad Miri - Hamid Mohammadnezhad

Computational Complexity of Sentiment Analysis Algorithms in Natural Language Processing

Kiana Karimifard - Mohammad Ghasemzadeh

Deep Learning-Based Prediction of Personality Profiles from Resting-State EEG Signals

Marjan Nosouhi - Moein Esghaei

Performance Analysis of Variational Quantum Classifiers for Classification Tasks in the NISQ Era

Saghar Kiani - Paria Kamalpour - Sarina Ghafouri - Niloofar Mirzaei Chahardeh

Strategies and Future Horizons of Innovative Entrepreneurship in AI-Based Programming

Milad Ghiasspour

Fibroglandular Tissue Classification in Breast MRI: A Comparative Study of Automated Decision Strategies

Meysam Khalaj - Arvin Arian - Ala Torabi - Nasrin Ahmadinejad - Masoumeh Gity - Seyedeh Nooshin Miratashi Yazdi - Mohammad Pooya Afshari - Melika Sadeghi Tabrizi - Hamid Soltanian-Zadeh

Generative AI and Individual Entrepreneurial Orientation

Saeed Mohammadi - Mahshid Mirzaee

بیشتر