Articles tagged with Computer Vision
Researchers have developed SMART-BARN, a cutting-edge technology lab for complex behavioral analysis. The facility can host hundreds of animals simultaneously and track their movements in 3D, allowing for unprecedented insights into collective behavior.
Researchers at Pohang University of Science & Technology have developed a sensor technology called computer vision-based optical strain (CVOS) that enhances durability and streamlines fabrication processes. This breakthrough enables the precise recognition of intricate bodily motions through a single sensor.
A team from Nanyang Technological University and the National University of Singapore aims to develop innovative solutions to enhance the accuracy of computer vision systems for autonomous vehicles. They will focus on designing CV systems that can recover to at least 80% of their original accuracy following physical threats or attacks.
Researchers developed a neural network-based system, PAT, for snapshot compressive imaging. It achieves comparable image quality to CASSI and holds strong promise due to advancements in AI processing capabilities.
Researchers developed a pair of modules to enhance polyp segmentation in colonoscopy images, overcoming challenges of image noise and camouflage. The new approach achieved significant improvements in accuracy, with a 2.6% increase in performance and an additional 1.8% gain from a camouflage detection module.
A breakthrough in photonic memory has been achieved, enabling fast volatile modulation and nonvolatile weight storage for rapid training of optical neural networks. The 5-bit photonic memory utilizes a low-loss PCM antimonite to achieve rapid response times and energy-efficient processing.
The Multi-frame Moving Object Detection System enhances remote sensing applications by detecting objects as small as one pixel in low-visibility conditions. It improves signal-to-noise ratio and detects fast- and slow-moving objects with high accuracy.
Researchers developed a deconvolution method for epidemiology using neural networks, inferring daily infection rates from mortality data. The approach can assess the effectiveness of non-pharmaceutical interventions like lockdowns and mask mandates in reducing infection transmission.
Researchers developed a fast and affordable test to predict cement durability using computer vision, which can analyze water droplet absorption on surfaces. The new test is less tedious than current methods and could help the cement industry improve quality control.
A new method developed by the Max Planck Institute of Animal Behavior has counted Africa's largest bat colony using GoPro cameras and artificial intelligence. The estimate puts the colony at between 750,000 and 1,000,000 bats, making it the largest for bats by biomass anywhere in the world.
Researchers at Carnegie Mellon University have created a new model, Vision-Robotics Bridge, which enables robots to learn from observing humans complete tasks in any environment. The robots successfully learned 12 tasks, including opening drawers and picking up objects, with minimal practice time of just 25 minutes.
Researchers have developed DyLiN and CoDyLiN, methods that handle non-rigid deformations and topological changes in 3D structure representation. These advancements enable real-time volumetric rendering and animation with improved visual fidelity and speed.
A new approach to enhance artificial intelligence-powered computer vision technologies has been developed by UCLA researchers, adding physics-based awareness to data-driven techniques. This hybrid methodology aims to improve how AI-based machinery sense, interact, and respond to their environment in real time.
Researchers at University of California - San Diego developed a new model that trains four-legged robots to see more clearly in 3D, allowing them to autonomously cross complex environments. The robot uses a forward-facing depth camera to synthesize visual information from past frames and estimate its surroundings.
MethaneMapper is an artificial intelligence-powered hyperspectral imaging tool that can detect real-time methane emissions and trace them to their sources. With a performance accuracy of 91%, it has the potential to revolutionize the way we monitor oil and gas operations and curb climate change.
A recent study published in Flora used social media images of cherry blossoms to track climate patterns and identify subtle off-season blooms. The researchers analyzed 10 years of data from Flickr and compared it with official records of cherry flowering times in Japan, finding a detailed seasonal pattern of blooming across the country.
Researchers at North Carolina State University have developed a new methodology called Patch-to-Cluster attention (PaCa) that addresses the challenges of vision transformers. PaCa improves ViT's ability to identify, classify, and segment objects in images while reducing computational demands and enhancing model interpretability.
A new MIT deep-learning system can analyze the internal structure and properties of materials based solely on their surface conditions. The technique uses vast amounts of simulated data to generate reliable predictions, offering a promising solution for engineers seeking non-invasive insights into material properties.
A team of IUPUI researchers has developed an AI-powered approach to classify insect species, tackling the challenge of discovering new species. The method uses deep hierarchical Bayesian learning to distinguish between known and unknown species, providing insight into their taxonomy and ecosystem impacts.
Researchers developed a robotic finger with high-resolution sensors that capture data along the entire length of each finger. The three-fingered robotic hand can identify objects after just one grasp, with 85% accuracy, using tactile sensing and machine-learning algorithms.
Researchers have developed an AI algorithm that can remove atmospheric blur from astronomical images, resulting in more accurate scientific measurements and clearer data. The tool produces faster and more realistic images than current methods, producing 38.6% less error compared to classic methods.
A new open-source tool called TILE2NET uses aerial imagery and image-recognition to create complete maps of sidewalks and crosswalks. The tool has been trained on 20,000 aerial images from Boston, Cambridge, New York City, and Washington, recognizing 90% or more of all sidewalks and crosswalks in these cities.
Researchers from University of Konstanz develop 'neural puppeteer' AI model to predict animal poses and appearances, enabling analysis of intermediate motions. The system uses 3D key points to calculate statistically likely steps, crucial for studying collective behavior in wildlife.
The UTSA ScooterLab will collect data on riders' mobility, context and environment to improve sustainable transportation solutions. The project aims to transform the way we think about micro-mobility.
Researchers at Columbia University are developing algorithms that enable robots to understand object permanence and learn from 3D information. This allows robots to track objects and humans as they move around, improving their perception capabilities in indoor environments.
Researchers developed an algorithm that uses computer vision techniques to estimate tree diameter from a single image in realistic field conditions. The app sped up the process significantly, being about four and a half times faster than manual measurement techniques.
Researchers employed computer vision to extract social behaviors and linked them to brain synchronization patterns in a novel approach. During cooperative play, brain synchronization was strongest when participants shared gaze, while individual play showed increased within-brain synchronization.
Scientists have developed AI techniques to track crevasses on the Thwaites Glacier, which could impact global sea levels by up to 60cm. The study found a complex interplay between crevasse formation and ice flow speed changes.
A pilot study conducted at Brigham and Women's Hospital found that a low-cost computer vision system was feasible and well-received by employees. The system accurately detected mask adherence 100% of the time, with most participants experiencing a positive interaction.
A novel multi-modal image retrieval system, DenseBert4Ret, has been developed by researchers from Gwangju Institute of Science and Technology (GIST) using deep learning algorithms. The system outperforms state-of-the-art models in retrieving images based on both image and text features.
Researchers used machine learning to track turbulent structures in fusion reactors, gaining detailed information on their behavior and heat flows. The approach enables more accurate engineering requirements for reactor walls and could lead to improved energy efficiency.
Researchers at MIT have developed a machine-learning model that captures how sounds propagate through spaces, allowing for accurate visual renderings of rooms. This technique has potential applications in virtual and augmented reality, as well as improving AI agents' understanding of their environment.
A team from the University of California San Diego has developed a new system of algorithms that enables four-legged robots to walk and run on challenging terrain while avoiding obstacles. The system combines vision with proprioception, allowing the robot to move efficiently and smoothly in various environments.
A study by Aston University confirms that taking a break every 20 minutes to look away from screens for at least 20 seconds reduces digital eye strain symptoms. The research, involving 29 participants, showed a marked decrease in symptoms such as dryness, sensitivity, and discomfort after using the reminders.
City digital twin technology is used to create synthetic training data for deep learning models, which are then trained on a combination of real and synthetic data. This approach yields promising results for architectural segmentation tasks, particularly for modern building styles.
Researchers at MIT have developed a machine-learning system that uses computer vision to monitor the 3D printing process and correct errors in real-time. The system successfully printed objects more accurately than other 3D printing controllers, enabling engineers to incorporate novel materials into their prints with ease.
Researchers developed a neural network algorithm that recognizes emotions and engagement from video images of faces, outperforming existing models in accuracy. The system can be integrated into video conferencing tools and online learning systems to analyze participant engagement and emotions.
A new robotic system, FuseBot, has been developed to efficiently retrieve buried objects in piles. The system uses radio frequency signals and computer vision to reason about the probable location and orientation of objects under the pile, enabling it to find more hidden items than a state-of-the-art robotics system in half the time.
Researchers at the University of Tokyo have developed a new method to detect deepfakes, using self-blended images that improve detection accuracy by 5-12%. The team created novel synthesized images with controlled artifacts to train algorithms and found significant improvements in detecting deepfake images and videos.
A new camera system developed by Carnegie Mellon University researchers can reconstruct sound vibrations with extraordinary accuracy, capturing isolated audio without inference or a microphone. The dual-shutter vibration-sensing system uses two cameras and a laser to detect high-speed, low-amplitude surface vibrations.
A team from KAUST has developed a low-cost system for imaging plant growth dynamics noninvasively and at high throughput. The Mutiple XL ab system combines computer vision and pattern recognition technologies with machine learning to analyze and quantify root growth dynamics.
Researchers at Carnegie Mellon University developed an AI-powered method for robots to recognize and pour transparent liquids like water. The technique uses contrastive learning for unpaired image-to-image translation, enabling robots to see through different backgrounds and pour accurately.
Snap Inc has endowed a professorship at TU Graz to develop visionary software methods in camera technology and explore new approaches to visual information processing. The professorship will pursue new application ideas for mixed reality, including the fusion of photos with computer-generated content.
Artificial Intelligence can now identify legendary batting techniques used by Sir Donald Bradman and modern players. Researchers developed a deep learning computer vision AI model to detect lateral backlift batters from straight ones.
Researchers at Penn State found emerging problems in Remote Sighted Assistance (RSA) technology that cannot be solved with existing computer vision techniques, warranting new development in human-AI collaboration. The technology connects visually impaired individuals with human agents for daily tasks requiring sight.
Researchers at Carnegie Mellon University developed AI-enhanced museum exhibits that increased learning and engagement for elementary school-aged children. The intelligent exhibits featured a virtual assistant, NoRilla, which interacted with visitors, asking questions and guiding them through scientific challenges.
New research suggests the brain uses multiple strategies to process smells, employing both snapshot-like and evolving ensemble approaches. The study provides new tools for scientists to quantify and interpret brain activity patterns.
MIT engineers mapped airplane contrails over the US in 2020 and found a 20% drop in coverage compared to prepandemic years. The team's computer-vision technique can help predict where contrails form, allowing airlines to reroute planes and reduce aviation's climate impact.
Adversarially robust models capture aspects of human peripheral processing, with results showing similarity in image transformations and perception alignment. The study's findings shed light on the goals of peripheral processing in humans and could help improve machine learning models.
KAUST researchers develop an artificial electronic retina that mimics human vision and recognizes handwritten numbers with high accuracy. The retina uses perovskite nanocrystals to detect light intensity via capacitive change, offering a more energy-efficient alternative to existing systems.
Researchers at Universidad Carlos III de Madrid developed a computer vision system to analyze cells in microscopy videos, allowing for automatic characterization of cell behavior. The system enables faster analysis of thousands of cells compared to traditional methods, which typically involve manual segmentation and tracking.
A team of scientists has developed a pioneering approach to combine advances in computer vision with ecological expertise to analyze wildlife populations. By leveraging AI and machine learning algorithms, researchers can extract key features from images and videos to quickly classify species, count individuals, and track behavior.
Researchers at the University of Groningen have developed an AI system that can recognize indoor spaces with high accuracy by combining image and audio data. The system achieved a 70% accuracy rate in recognizing nine different types of indoor spaces, surpassing previous results.
Researchers from KTU proposed a deep-learning-based method for 3D human shape reconstruction using limited-angle depth data. The method can be integrated with existing virtual reality tools and has potential applications in telemedicine and remote diagnostics.
Researchers developed a new hand gesture recognition algorithm that surpasses current methods in accuracy, complexity, and applicability. The algorithm combines adaptive hand type classification and a shortcut feature for efficient real-time recognition.
A team of biologists and engineers created a robotic fish that scares mosquitofish away, altering its behavior and physiology. The study found that the mosquitofish showed fearful behaviors, weight loss, and reduced fertility when confronted with the robot.
A recent study used computer vision algorithms to analyze nearly 9,400 Flickr photos taken along Colorado's Front Range, identifying preferred outdoor landscapes with moderate accuracy. The algorithm performed well for images of water, structures, and agricultural lands, but struggled with forests. Combining social media data with on-s...
Researchers developed an automated system using deep learning to detect COVID-19 lesions in CT chest scans, achieving 99% accuracy. The system can provide high-precision data for doctors to make robust and accurate diagnoses.
Researchers at MIT develop RFusion, a robotic system that uses data from a camera and radio frequency antenna to locate and retrieve lost items. The system relies on RFID tags and machine learning algorithms to optimize the robot's trajectory and grasp the object.
The Imageomics Institute, led by The Ohio State University, aims to use machine learning methodologies to extract biological traits from images of living organisms. Experts like Chuck Stewart will utilize computer vision and artificial intelligence to help infer phylogenetic traits from images.