Abstract: Deepfake technology, driven by generative models such as GANs and diffusion architectures, has enabled the creation of highly realistic manipulated media capable of deceiving both visual and auditory perception. Such forgeries pose significant risks to identity verification, digital forensics, and public trust. Existing detection methods are often restricted to single modalities or handcrafted features, limiting their robustness against high-quality cross-modal deepfakes. This work presents a multi-modal deepfake detection framework that integrates visual and audio analysis. The system employs a ResNet-based backbone for spatial feature extraction, Bi-LSTM encoders for temporal modeling in videos, and a 2D CNN with Random Forest classification for speech-based anomaly detection. An attention-based late-fusion strategy combines outputs to generate reliable predictions with calibrated confidence scores. Experimental results demonstrate high performance, with training accuracy reaching 98% and validation accuracy of 92% on benchmark datasets. The system is deployed as a Flask-based web application, enabling real-time detection and interpretable outputs.
Background and Objective: Self-harm is a psychologically damaging behavior, and its accurate differentiation from other wounds (violence, accidents, burns, diabetic ulcers) is critically important in forensic medicine. However, this differentiation often falls into a diagnostic "…
This repository contains the supplementary materials associated with the article: “Beyond Grades: Multi-Target Deep Learning for Early Academic Risk Detection” The materials support the transparency, reproducibility, interpretability, and pedagogical analysis of the leakage-free…
The cardiovascular (Cardiac) disease (CVD) is another factor that causes death among the global population most, and this is the reason why there is a high necessity to implement proper, effective, and interpretive diagnostic systems. The usage of machine learning (ML), deep lear…
A hybrid deep learning framework for monthly precipitation prediction in mountainous areas of Boyacá, Colombia. Combines Graph Neural Networks (GNN) with temporal attention mechanisms and ConvLSTM architectures for accurate spatiotemporal forecasting. This implementation includes…
Self-supervised deep learning has emerged as a powerful method for image enhancement when a priori ground-truth references are not available. Stemming from Noise2Noise , it was shown that a convolutional neural network (CNN) can be trained from a noisy input and target pair of th…
## ALTERNATIVE TITLES ### Alternative Title 1 (Comprehensive)**"AI-Driven Analysis of Cube {100}<001> and Goss {110}<001> Textures: Machine Learning, Deep Learning, and Generative Models for Crystallographic Texture Quantification in Metallurgical Engineering"** ### Alternative T…