Artificial Intelligence

Traditional authentication models are vulnerable to security breaches when personal data is exposed. This study introduces novel hybrid visual stimuli protocols integrating event-related potentials (ERP) and steady-state visually evoked potentials (SSVEP) to develop an authentication system that enhances both performance and personalization in neural interfaces. Our model utilizes distinctive neural patterns elicited by a range of visual stimuli based on 4-digit numbers, such as familiar numbers (personal birthdates, excluding targets), standard targets, and non-targets.

Traditional Thai medicine (TTM) is an increasingly popular treatment option. Tongue diagnosis is a highly efficient method for determining overall health, practiced by TTM practitioners. However, the diagnosis naturally varies depending on the practitioner's expertise. In this work, we propose tongue image analysis using raw pixels and artificial intelligence (AI) to support TTM diagnoses. The target classification of Tri-Dhat consists of three classes: Vata, Pitta, and Kapha. We utilize our own organized genuine datasets collected from our university's TTM hospital.

Existing datasets of infrared and visible images only contain few extreme scenes, we construct a dataset of images with haze based on the M3FD dataset. We pick 450 aligned image pairs from M3FD dataset and synthesize hazy visible images using the ASM. Due to the unique imaging principle of infrared images, rarely affected by haze, there is no need to do additional process for infrared images. Finally, a dataset named MHS has been released, which contains 450 pairs of images in hazy conditions.

Current neural network solutions for channel estimation are frequently tested by training and testing on one example channel or similar channels. However, data-driven algorithms often degrade significantly on other channels which they are not trained on, because they cannot extrapolate their training knowledge. Online training can fine-tune the offline-trained neural networks to compensate for this degradation, but its feasibility is challenged by the tremendous computational resources required.

We introduce an online-offline Iraquian hand-drawing dataset for early Parkinson’s disease detection, exclusively collected using smartphones, thus eliminating the need for specialized equipment like digitizing tablets and pens. Our dataset comprises data from 30 healthy individuals (17 men, 13 women) with an average age of 56 years (SD = 6.12) and 30 PD patients (23 men, 7 women) with an average age of 60 years (SD = 4.91), gathered at Marjan Hospital in Hilla, Babil Governorate, Iraq.

SETCD (Satellite and ERA5-based Tropical Cyclone Dataset), a comprehensive dataset encompassing satellite imagery data and ERA5 data for all TCs recorded between 1980 and 2022. Our dataset is derived from two publicly available data sources: GridSat-B1 and ERA5. To capture relevant information associated with TC, SETCD adopts the latitude and longitude positions provided by IBTrACS as the center points. The satellite data within the SETCD dataset consists of three channels from GridSat-B1: infrared, water vapor, and visible.

KPI prediction, which is categorized under time series data modeling, serves as a crucial area of investigation within the realm of complex industrial processes. This field focuses on forecasting key performance indicators that are pivotal for assessing the operational efficiency and productivity of industries. By leveraging historical data trends, KPI prediction aids in optimizing process controls and decision-making strategies, thus enhancing overall performance and competitive edge.

The WPT dataset was specially created for "Web Page Tampering Detection Based on Dynamic Temporal Graph Pre-training" and encompasses over 200,000 regular web pages from 75 websites across the finance, healthcare, and education sectors, in addition to 1,541 tampered examples sourced from zone-h.org. This dataset organizes web pages as nodes and their links as edges within a discrete dynamic graph structure, capturing snapshots at various moments in time. Each node integrates structural, textual, and statistical features into a robust 148-dimensional feature vector for every page.

Weibo and Twitter

1)The Weibo dataset is derived from the Weibo social platform. The collection of true information in this dataset originates from authoritative Chinese sources, while fake information is acquired through the official Weibo rumor suppression system. Each data instance within this dataset comprises both a news text and a corresponding news image.