late-fusion

Baseline model for multimodal classification based on images and text. Text representation obtained from pretrained BERT base model and image representation obtained from VGG16 pretrained model.

Visual Fusion of Camera and LiDAR Sensor

Research software and privacy-safe aggregate results for multimodal classification of primary liver tumors on multiphase CT, with internal fusion evaluation and external transportability stress testing.

This study introduces MultiBanFakeDetect, a novel multimodal dataset for Bangla fake news detection, combining textual and visual information. It features TextFakeNet for text analysis and MultiFusionFake for integrating multimodal data.

🛡️ Multimodal deepfake detection & explainable AI digital forensics workstation. Features video/audio late-fusion classifiers, Grad-CAM visual heatmaps, and an offline forensics RAG agent.

Early Fusion, Late Fusion, and Hybrid Fusion

🧠 Recognize emotions from text, audio, and video to generate adaptive CBT-style responses for digital mental health support.

This project aims to develop multimodal deep learning model for fall detecting(Sleep or Fall)

Modular training and evaluation framework for semantic segmentation of aerial and Sentinel-2 satellite imagery (FLAIR-2 dataset)

Implementation of the paper "Fusion Strategies for Breast Cancer Characterization Using Traditional and Deep Learning Models".

Auditable Victoria 3 GraphRAG mapping player slang to official mechanics with concept graphs, dual-view retrieval, MCP tools, and traceable evidence.

Multimodal deepfake detection (video + audio + lip-sync) on FakeAVCeleb v1.2 — Xception + Wav2Vec2 + sync stream, late fusion, multi-task heads (any-AUC 0.9994).

Late-fusion PyTorch pipeline for football match prediction — LSTM over rolling match stats + xlm-roberta text embeddings, with chronological splits

RAG Evaluation project with early, late, and hybrid fusion strategies

Serrano Team at EXIST 2026 Task 2: hierarchical multimodal sexism analysis for EN/ES memes, with audited official results and a reproducible baseline.

🎧 Ensemble multimodal de deteção de frustração para handover automático em apoio ao cliente — MIA, UMinho 2025/26

Research prototype for leakage-aware medical-image classification, clinical document and typed-text analysis, and same-study IU chest X-ray–Indication late fusion with template-based Findings generation.

A Dual-Stream Multimodal Fusion Framework for Breast Cancer Detection

🧠 Mental Health Risk Prediction: A machine learning-based system designed to predict mental health risk levels (Low, Medium, High) using survey data from multiple countries. The project includes data preprocessing, EDA, clustering, feature selection, and classification models.

Applied multimodal ML system combining text, audio, and visual emotion recognition with CBT-aligned response generation using late fusion.