Articles tagged with Deep Learning
Recent high-quality deepfake videos can feature realistic heartbeats and minute changes in face color, making them challenging to detect. Researchers found that even small variations in skin tone and facial motion can replicate the original pulse in deepfakes.
Researchers from OIST and Hebrew University developed a novel method to measure energy usage during movement using video and 3D-tracking via deep learning. This innovative approach expands the study of movement energy in ecology, physiology, and beyond, enabling the accurate measurement of energy consumption in smaller animal species.
A new AI tool has been developed to predict the relapse of pediatric brain cancer with high accuracy, using temporal learning algorithms to analyze sequential brain scans. The tool achieved an accuracy of 75-89% in predicting recurrence, outperforming predictions based on single images.
Researchers developed an IEAC framework combining robust security with high-capacity transmission performance, achieving a record 1 Tb/s secure transmission over 1,200 km of optical fibre. The system eliminates the trade-off between security and speed by integrating encryption into the communication process.
A Lehigh University team developed a novel machine learning method to predict abnormal grain growth in materials, enabling the creation of stronger, more reliable materials. The model successfully predicted abnormal grain growth in 86% of cases, with predictions made up to 20% of the material's lifetime.
Dr. Latifur Khan, a renowned computer science professor, has been elected as an AAAS Fellow for his pioneering work in machine learning and big-data analytics. He developed innovative solutions to adapt machine learning models to cybersecurity risks and created an AI-driven tool to analyze political conflict and violence.
A new hardware platform for AI accelerators capable of handling significant workloads with reduced energy requirement has been developed. The platform leverages III-V compound semiconductors to create photonic integrated circuits, which operate at the speed of light with minimal energy loss.
Researchers developed a method to detect epileptic seizures in humans using canine EEG data. The approach leverages feature similarities across species and modalities, reducing input space discrepancies. Euclidean alignment and knowledge distillation are key components of the proposed joint alignment mechanism.
Researchers at Concordia University have developed a new approach to identifying fake news on social media using the SmoothDetector model. The model integrates probabilistic algorithms with deep neural networks to capture uncertainties and patterns in multimodal data, providing more nuanced judgments of authenticity.
Researchers have introduced Orion, a novel framework that brings fully homomorphic encryption to deep learning, enabling computations on encrypted data without decrypting it. The framework achieves a 2.38x speedup over existing state-of-the-art methods and enables high-resolution FHE object detection using large neural networks.
Researchers developed an AI model that classifies variable stars from light curves with high accuracy, outperforming traditional approaches. The StarWhisper LightCurve series achieves near 90% accuracy with minimal manual intervention, paving the way for parallel data analysis and multi-modal AI applications in astronomy.
The study systematically traces Generative AI evolution from deep learning to foundation models, highlighting four distinct stages and successful applications. Key challenges like safety concerns and theoretical breakthroughs require further attention and development in the field of Generative AI.
Researchers from Osaka Metropolitan University used a deep learning model to discover new bubble-like structures in the Milky Way galaxy, providing insights into star formation and galaxy evolution. The study also revealed shell-like structures formed by supernova explosions.
A research team at Kumamoto University developed a deep learning-based method for analyzing the cytoskeleton more accurately and efficiently than ever before. This technique enabled more reliable measurements of cytoskeleton density, which is critical for understanding cellular structure and function.
A new AI model developed by UC Riverside scholars combines historical sales data with economic demand theory to predict prices in uncertain times. The model retains high accuracy, demonstrating a substantial improvement over other methods in reducing generalization errors.
Researchers developed a deep learning algorithm to denoise ultra-low dose CT scans, improving image quality and accuracy. The study found that this approach can diagnose pneumonia in immunocompromised patients using only 2% of the radiation dose of standard CT scans.
Researchers developed single-shot super-resolved fringe projection profilometry (SSSR-FPP) using deep learning to achieve 100,000 frames-per-second 3D imaging. This breakthrough offers new insights into ultra-fast dynamic processes and could revolutionize fields like mechanics and biology.
A new study suggests that artificial intelligence can effectively detect wildfires in the Amazon rainforest, using satellite imaging and deep learning. The technology achieved a 93% success rate in training models via datasets of images with and without wildfires.
This study utilized deep learning models to diagnose and predict the likelihood of malignant transformation in oral potentially malignant disorders. AI-driven approaches offer noninvasive, cost-effective, and objective means to enhance early detection and improve patient outcomes.
A deep learning model, CGMformer, leverages large-scale continuous glucose monitoring (CGM) data to extract individual glucose dynamics. The model captures a continuous picture of glucose fluctuations, identifying patterns that may indicate early metabolic dysfunction.
A new deep learning approach called HikingTTE significantly improves hiking travel time estimation by considering individual walking ability and fatigue. The model outperformed conventional techniques, reducing Mean Absolute Percentage Error by 12.95 percentage points.
The National Center for Supercomputing Applications has been awarded a grant to continue its Research Experience for Undergraduates (REU) program, which provides students with hands-on experience in machine learning and deep learning projects. The program aims to develop open source machine learning models and tools and apply them to r...
Deep Nanometry enables high-speed analysis of nanoparticles, detecting even trace amounts of rare particles like extracellular vesicles indicative of colon cancer. This technique has potential applications in various fields including vaccine development and environmental monitoring.
Researchers developed an innovative model, ECG-LM, that leverages large language models to interpret complex ECG signals. The model improves the accuracy and speed of heart-related diagnostics, particularly in resource-limited environments.
A team of researchers from Sophia University demonstrated the potential of combining social media posts and transformer-based learning models to detect heat stroke risks. The study found that a Japanese model called LUKE achieved high accuracy, suggesting its viability in monitoring heat stroke risks during heat waves.
Researchers have created a geometric deep learning approach called MARBLE that can infer latent brain activity patterns across experimental subjects. The method uses dynamic motifs to analyze neural population recordings and has been shown to be more interpretable than other machine learning methods.
Researchers used deep learning models to compare gene regulation in different cell types of human and chicken brains, shedding new light on brain evolution and providing tools for studying gene regulation. The study found that while some cell types are highly conserved between birds and mammals, others have evolved differently.
Researchers develop a new framework called MSF-Net to improve WiFi-based human activity recognition, achieving high accuracy scores compared to state-of-the-art techniques. The technology has potential applications in smart homes, rehabilitation medicine, and care for the elderly.
A new deep learning technique called CTLESS enhances myocardial perfusion imaging accuracy without requiring additional radiation scans. This method leverages deep learning to estimate attenuation maps, improving diagnostic interpretation and potentially boosting technological health equality across the U.S. and worldwide.
Researchers have proposed a new fault-tolerant framework for distributed deep learning model training that minimizes overhead and improves efficiency. By utilizing idle system resources during training, the framework effectively coordinates tasks with fault-tolerance functions.
A recent study introduces an innovative method for analyzing body composition, providing accurate assessments of body fat and muscle distribution. The approach utilizes deep, nonlinear methods to enhance estimation accuracy, surpassing previous linear models.
A recent study by University of New Hampshire researchers explores how to standardize rock climbing route difficulty through machine learning techniques, aiming to promote inclusivity and accuracy. The most successful approach used route-centric natural language processing methods, achieving an accuracy of 84.7%.
A new deep learning model has been developed to detect and segment lung tumors on CT scans with high accuracy. The model achieved 92% sensitivity and 82% specificity in detecting lung tumors, outperforming physician-delineated volumes in some cases. However, further research is needed to improve the model's performance for larger tumors.
Researchers at Sharjah University have designed techniques to automatically predict suitable image dimensions using deep learning models. The proposed methods aim to bridge the gap in automating image retargeting approaches based on an image and target resolution.
The EPFL team has developed a deep-learning pipeline called MaSIF to design new proteins that interact with therapeutic targets. They have successfully designed novel protein binders that can recognize and bind to drug-protein complexes, offering potential applications in cell-based therapies and biosensors.
A new 6D pose dataset has been introduced, providing high-quality RGB and depth images with annotated 6D pose data. This dataset achieves state-of-the-art accuracy rates of 97.05% and 98.09% for robotic grasping and automation applications.
Researchers developed an AI-powered technology that transforms low-resolution, label-free images into high-resolution, virtually stained ones without fluorescent dyes. This innovation delivers stable and accurate cell visualization, overcoming limitations of traditional imaging methods.
Researchers at the University of Washington have developed new proteins that can neutralize lethal snake venom toxins using deep learning computational methods. These protein designs show promise for creating safer and more cost-effective antivenoms, potentially saving millions of lives annually.
Researchers have developed a new geometric machine learning method called MaSIF, which enables the design of proteins that bind specifically to desired molecular structures. This approach accelerates precision drug development by allowing for precise dosing and control of biological drugs.
Researchers developed an explainable deep learning model to predict and analyze HABs in Chinese lakes and reservoirs, achieving significant improvement over conventional machine learning methods. The model identified water temperature as the most influential factor driving algal bloom dynamics.
A novel method called RESQUE predicts computational and energy costs for updating deep learning/AI models, allowing users to make informed decisions about when to update models to improve AI sustainability. The researchers conducted extensive experiments to validate the performance of RESQUE.
Researchers developed TLE-PINN to predict melt pool morphology in selective laser melting, achieving superior accuracy and faster training times. The framework combines physics-informed constraints with deep learning techniques, enabling precise and efficient solutions for real-time process control and manufacturing optimization.
The one-core-neuron system (OCNS) minimizes model size while maintaining high performance in deep learning. OCNS employs a single neuron to encode high-dimensional data into a one-dimensional time-series representation, reducing parameters by an average of 0.035%.
The SSIT model uses a single encoder to extract spatial features and a decoder to reconstruct images with desired content and style. It outperforms other GAN models in image transformation tasks, offering potential for democratizing image transformation on devices like smartphones.
Researchers developed a deep learning model that classifies pancreatic cancer into molecular subtypes using histopathology images, achieving high accuracy and rapid turnaround time. The AI tool has the potential to improve patient outcomes by enabling timely and tailored treatment strategies.
A deep learning model developed by researchers at San Diego State University can accurately diagnose chronic obstructive pulmonary disease (COPD) using a single inhalation lung CT scan. The study found that the model performed similarly to traditional two-phase CT measurements, with added clinical data improving accuracy.
Researchers have developed an algorithm-based scheme to help drivers avert drowsiness, which contributes to thousands of fatal incidents and injuries every year. The system uses EEG signal detection and machine learning algorithms to achieve high accuracy and reduce training time.
A new computational model called Multi-Stage Residual-BCR Net (m-rBCR) uses a unique frequency representation to solve deconvolution tasks with fewer parameters and faster processing times. The model demonstrates high performance on various microscopy datasets, outperforming traditional methods.
Researchers have developed a deep-learning-powered metalens imaging system that overcomes limitations of traditional metalenses. The system pairs a mass-produced metalens with an image restoration framework driven by AI to achieve aberration-free, full-color images while maintaining compact form factor.
Researchers developed a deep learning-based method for identifying 2D materials using Raman spectroscopy, achieving high classification accuracy and reducing manual intervention. The new approach generates synthetic data to enhance datasets, enabling precise material characterization even with scarce experimental data.
A deep learning AI model can identify pathology in images of animal and human tissue much faster and often more accurately than people, potentially revolutionizing disease-related research and medical diagnosis. The model was trained using images from past epigenetic studies and showed accuracy comparable to human experts.
The study explores gene fusion technologies, including FISH, PCR, IHC, ECL, and NGS, to detect biomarkers in tumor diagnosis. AI-driven detection and comprehensive genome-wide analysis using NGS and bioinformatic tools enhance diagnostic accuracy.
Researchers introduce a novel approach to multiplexed fringe projection profilometry using deep learning and frequency-domain multiplexing. This method achieves high-resolution and high-speed 3D imaging at near-one-order of magnitude-higher frame rates with conventional low-speed cameras.
Researchers developed a novel noninvasive choroidal angiography method using deep learning, enabling layer-wise visualization and evaluation of choroidal vessels. The approach employs an advanced segmentation model to handle varying quality of OCT B-scans, offering a promising tool for clinical applications.
Researchers are using deep learning to help protect chimpanzees in the Greater Mahale Ecosystem, Tanzania. A new acoustic detector has been developed to identify chimpanzee sounds and monitor population density more efficiently, allowing for better conservation strategies.
A new study from Virginia Tech shows large language models can assess human-made environments using street-view images, similar to traditional methods. LLM-based performance offers a more accessible tool for users in small to medium-sized cities, making it easier to manage smart urban infrastructure.
A new review article explores the transformative role of deep learning techniques in revolutionizing protein structure prediction. Deep learning models like AlphaFold 2 have achieved high accuracy, over 98%, in predicting human protein structures, surpassing traditional methods.
A novel collaborative framework integrates semi-supervised learning techniques to improve MRI segmentation accuracy, even with limited labeled data. The approach achieves high Dice scores and demonstrates its potential for practical clinical application.
A new study finds that AI-powered models exhibit similar levels of accuracy as ophthalmologists in identifying infectious keratitis, a leading cause of corneal blindness worldwide. The AI models displayed a sensitivity and specificity of 89.2% and 93.2%, respectively, matching the diagnostic accuracy of human experts.
Deep learning models used in remote sensing tasks are susceptible to various types of noise and attacks, compromising their performance. The study assesses the vulnerabilities of DL algorithms for object detection, revealing several weaknesses that can be leveraged by attackers.