Articles tagged with Computer Vision
Researchers at Aalto University developed an augmented climbing wall, combining body tracking and custom software to empower users as content creators. The system offers diverse movements, challenges, and endless gaming experiences, increasing the sport's appeal to new audiences.
A University of Rochester team has developed a system that outperforms other approaches to creating computer-generated image captions by considering the meaning and context of words, not just images. The winning approach combines top-down and bottom-up methods to create more accurate and coherent captions.
Researchers found that humans can recognize objects in partial- or low-resolution images, with a
Researchers at NYU Langone Health found that non-native English speakers performed slower on sideline vision tests and had higher saccade rates compared to native English speakers. The study suggests that clinicians and trainers need to consider language when interpreting test results, which may impact concussion detection.
Disney researchers have developed a method to estimate pose and predict orientation of objects, using similarities in how different types of objects appear from the same angle. The system proved effective in predicting pose even for unseen objects, with applications in self-driving cars and other computer vision tasks.
A new algorithm combines traditional computer vision classification with deep learning models to improve pedestrian detection speed and accuracy. The technology has the potential to be used in smart vehicles, robotics, and image/video search systems.
A recent study by scientists from the University of Exeter and Cambridge has confirmed that camouflage plays a crucial role in protecting animals from predators. The research found that animals or eggs with matching patterns or contrasts to their surroundings were less likely to be eaten.
A Dartmouth-Penn study finds that while male-dominated, the gender gap is closing in vision science, with a substantial generational difference in gender balance. The results suggest that younger generations have a smaller gap, but women still face higher dropout rates and underrepresentation in recognitions.
Researchers at Newcastle University have confirmed that praying mantises use 3D vision to hunt, with a new model to improve visual perception in robots. The team used custom-made glasses with one blue and one green lens to show insects any desired images.
The FaceDirector system enables directors to fine-tune performances in post-production, saving time and money by avoiding reshoots. It combines facial expressions and audio cues for optimal synchronization, allowing users to generate novel versions of performances.
Researchers developed a new technique called photogeometric scene flow (PGSF) that combines three computer vision methods to capture high-quality and detailed facial features. The method produces superior results in capturing facial details, making it extremely valuable for realistic facial reconstructions.
University of Washington researchers have developed a technology to capture the 'persona' of a well-photographed person like Tom Hanks from vast numbers of Internet images. The digital model can be animated to deliver speeches that the real actor never performed, and even transfer expressions and mannerisms onto another person's face.
Researchers at Carnegie Mellon University are building a wearable cognitive assistance system called Gabriel that provides instructions for tasks like repairing equipment or assembling furniture. The system uses a wearable vision system and taps into cloud computing via 'cloudlets' to enable real-time guidance.
A research group at Disney Research Pittsburgh developed a computer vision system that continuously improves its ability to recognize objects by picking up hints from videos. The system outperformed other methods in detecting various objects, including microwave ovens and stoves.
The Merlin Bird Photo ID system can identify 400 bird species in the US and Canada with accuracy of 90% and is designed to improve with user input. The system combines AI techniques with millions of data points from humans to present the most likely species, including photos and sounds.
Computer scientists developed a new method combining computer vision algorithms and brain-computer interface to detect mines in sonar images, outperforming existing methods. The system uses classifiers to capture changes in pixel intensity and improves accuracy by detecting 99.5% of true positives and reducing false positives.
A new computerized vision-screening test, the Jaeb Visual Acuity Screener (JVAS), has been developed to identify children with subnormal visual acuity. The test uses a set testing algorithm to minimize subjective tester bias and provides simple pass/fail results for four age groups.
A new system designed by researchers from Brown and Johns Hopkins universities aims to assess computer vision systems' ability to understand the context of an image. The 'visual Turing test' evaluates how well computers can recognize subtle details, such as people walking together and having a conversation.
Virginia Tech researcher Devi Parikh aims to build intelligent machines that can understand the visual world from images and videos. She proposes teaching computers through visual abstractions, such as cartoons, which she believes are easier to illustrate than complex concepts.
Researchers at UC Berkeley are developing a vision-correcting display that uses computation to compensate for individual visual impairments. The technology has the potential to transform lives of people with high-order aberrations and presbyopia, enabling them to use smartphones, tablets, and computers without corrective lenses.
The new Birdsnap app, developed by Columbia Engineering researchers, can identify 500 common North American bird species using computer vision and machine learning techniques. It offers users various ways to organize species and even annotates images with distinctive parts for easy identification.
A new activity-recognition algorithm has been developed, enabling computers to efficiently search video for actions. The algorithm's execution time scales linearly with the size of the video file, and it can make good guesses about partially completed actions.
A new algorithm developed by MIT researchers can aid robots in navigating unfamiliar buildings and understanding scenes. The algorithm identifies dominant orientations in 3D scenes, making it easier to re-identify landmarks and segment planes.
Researchers found that vision restoration training can strengthen partially surviving neurons and promote visual recovery 'hot spots'. The study analyzed data from 32 stroke patients with hemianopia and found that average absolute improvement was 6%.
The Never Ending Image Learner (NEIL) program analyzes 3 million images, identifying 1,500 object types and 2,500 associations, and develops a growing visual database to enhance computer vision capabilities. NEIL's findings are made available online, providing insights into common sense knowledge that humans take for granted.
Christian Theobalt aims to enable computers to reconstruct motions and surface characteristics from video camera input. His project, CapReal, will explore theoretical bases for whole new methods of dynamic scene reconstruction, aiming to capture geometry, motions, and material characteristics in complex real scenes.
Researchers at MIT have developed a system called Wi-Vi that uses low-cost Wi-Fi technology to track human movement through walls and closed doors. The system cancels out reflections from static objects, allowing it to detect only moving humans.
Researchers at Carnegie Mellon University developed a process called Lifelong Robotic Object Discovery (LROD) that enables a two-armed robot to discover objects using color video, Kinect depth camera, and non-visual information. The robot can refine its understanding of objects over time, focusing on those most relevant to its goal.
A new study published in the Journal of Vision increases our understanding of how the brain processes facial structure and recognizes family resemblance. Researchers found that people can pick out family members despite underlying differences, such as gender or age, by comparing faces to an average face for that gender.
Researchers at Tel Aviv University have developed software using facial recognition technology to identify and join digitized fragments of the Cairo Genizah collection. This has led to the discovery of pages from a work by Saadia Gaon, a prominent rabbi and philosopher from the 10th century.
Scientists at IVIA have created a machine that detects rotten oranges using computer vision. Another prototype classifies mandarin segments by quality and damage. These machines improve efficiency in the fruit selection process.
Researchers at the University of Washington have developed a tool that can take hundreds of digital portraits and create an animation of a person's face in seconds. The software aligns faces and chooses photos with similar expressions to produce smooth transitions, allowing users to see how a person's expression changes over time.
Researchers develop two new techniques that apply heat mapping and heat distribution to enable machines to perceive three-dimensional objects. These methods mimic the human ability to recognize objects regardless of their shape or orientation.
Researchers at New York University's Courant Institute of Mathematical Sciences developed an algorithm for pose estimation, which involves detecting specific objects or individuals under various conditions. They achieved this by utilizing a unique visual database from the Dutch band C-Mon & Kypski's music video crowd-sourcing project.
The NYU Courant Institute is developing a bird-sized, self-flying plane with AI-powered vision for navigation. The plane will use convolutional networks to quickly interpret its environment and navigate through forests and urban environments.
Takeo Kanade has been honored with the ACM/AAAI Allen Newell Award for his career contributions to computer vision and robotics. He is being recognized for bridging computer science and other disciplines through his groundbreaking work, including the development of face recognition systems and robotic technologies.
A new study by MIT neuroscientists suggests that the brain learns to solve object recognition through experience in the natural world. Researchers exposed monkeys to altered visual stimuli, which led to changes in neural activity and object preferences.
The COGNITO project, led by the University of Bristol, aims to develop computer systems that can understand and predict human behavior and activity. By using on-body sensor networks and advanced computer vision algorithms, researchers hope to create systems that can assist people in skilled assembly and manipulation tasks.
A new method devised by Carnegie Mellon researchers enables computers to gain a deeper understanding of an image by reasoning about physical constraints. The method uses virtual blocks to build a three-dimensional approximation of the scene, improving accuracy in estimating layout and surface details.
Researchers developed a proof-of-concept system that removes pedestrians from urban scenes in Google Street View, exploring how computer vision can preserve privacy in public environments. The system replaces pedestrians with an approximation of the actual background behind each pedestrian.
Babenko's fellowship will allow him to attend the European Conference on Computer Vision and work on weakly supervised learning systems for object detection and recognition. His research focuses on tackling computer vision problems with less precise labels, enabling more efficient data collection.
A new study suggests that people with migraines may struggle to filter out visual noise, even when they don't have a headache. Migraineurs performed significantly worse in identifying visual targets than those without migraines when visual distractions were added.
Researchers from North Carolina State University have developed a computer program that enables cars to stay within their lane, opening the door to new safety features and military applications. This technology can detect traffic and lanes, allowing cars to avoid accidents caused by driver fatigue or distraction.
Researchers at Carnegie Mellon University's National Robotics Engineering Center developed a plant-sorting machine that uses computer vision and machine learning to classify and sort harvested strawberry plants more consistently and faster than human workers. The system has the potential to improve quality, streamline production and de...
Researchers from Harvard and MIT demonstrate a new method to build better artificial visual systems by combining genetic screening techniques with high-performance gaming hardware. The approach enables the analysis of thousands of candidate models in parallel, resulting in significant improvements over traditional methods.
A research study is being conducted to assess the safety and effectiveness of an investigational corneal inlay in reducing the need for reading glasses. The inlay, called AcuFocus Corneal Inlay (ACI), is intended to improve near vision in patients with presbyopia, a natural loss of near vision that occurs after age 45.
Researchers developed a technique inspired by the human eye to improve object recognition, surveillance, and shape reconstruction. The new algorithm uses trust search regions to quickly identify objects with nearly double the accuracy of previous methods.
A study published in the American Journal of Preventive Medicine found a more-than-sevenfold increase in computer-related injuries from 1994 to 2006. The majority of injuries occurred at home and were caused by tripping or falling over equipment, hitting against or catching on computers, or computer equipment falling on patients.
Computer scientists at UC San Diego have created a software program that can duplicate keys from photographs, highlighting the vulnerability of key-based security systems. The system uses computer vision techniques to extract the numeric code from the key's bumps and valleys, making it possible to create identical copies.
A new study found that office-based vision therapy by a trained therapist, along with at-home reinforcement, is more effective in treating convergence insufficiency (CI) in children than traditional home-based therapy. The treatment resulted in reduced symptoms and improved vision in nearly 75% of participants.
Researchers found that our brains use the timing of visual input to recognize objects, even when they appear differently. By analyzing neural activity in monkeys, they demonstrated that the brain can learn to confuse similar images through temporal contiguity, leading to improved object recognition.
Researchers analyzed ancient magic tricks to gain insights into human psychology, revealing the power of misdirection and illusion in shaping our perceptions. The study's findings have long-term implications for fields like human-computer interfaces, where understanding attention and perception is crucial.
Researchers at Carnegie Mellon University developed an algorithm that can geolocate photos by analyzing their composition and searching through millions of GPS-tagged images on Flickr. The IM2GPS algorithm achieved accurate results, even identifying landmarks like the Eiffel Tower and narrowing down possibilities like beach or desert l...
The UC-San Diego computer scientists have developed an algorithm to separate gloss from non-gloss in digital images, enabling the creation of photorealistic makeup portraits. Users can upload a portrait-style photograph and apply thousands of makeup products from various brands.
A team of UC San Diego computer scientists has developed a groundbreaking algorithm for 3D reconstruction from 2-D images, solving the long-standing problem of autocalibration. This technology has vast potential in various applications, including virtual product showcases and augmented-reality walkthroughs.
A recent study by MIT researcher Tomaso Poggio found that a computer model was able to reproduce human behavior in recognizing complex scenes, leading to a significant shift in understanding the visual cortex. The findings have implications for understanding how the brain works and may lead to advancements in artificial intelligence.
Researchers at Frankfurt's Research Center for Neurotechnology have developed new artificial vision systems using neural organization principles. These systems aim to combine individual functions to form a functioning whole, leading to potential applications in intelligent security and autonomous vehicles.
Researchers at the University of Washington have created 3D models of landmarks using vacation photos from Flickr. The models were accurate enough to show individual figures carved into stone facades, with resolutions comparable to those obtained by laser scanners.
A computer-based method to evaluate the eye's ability to distinguish object details and shape in patients with AMD could provide a more accurate way to assess the effectiveness of eye surgery or vision rehabilitation interventions. The test might help detect early detection of AMD and estimate patient potential visual acuity.
The USC team aims to design visual displays that enhance certain parts of an image to help people with AMD digest visual information better. They will use techniques from computer vision and computational neuroscience to build visual displays that compensate for neural loss.