Researchers have developed a novel audio-visual model that can isolate and enhance speech in videos, even in challenging real-world scenarios. The model uses both visual cues, such as lip movements, and auditory signals to focus on the speaker's voice and improve speech quality.
Researchers have developed a novel method enabling infinite walking in VR by manipulating user eye movements during saccadic suppression. The system successfully redirects users away from obstacles in the physical space while avoiding scene distortions.
Researchers at Aalto University developed a new computational method to address the challenge of computing realistic indirect illumination in real-time computer graphics. The novel method accurately computes indirect illumination using sparse radiance probes, enabling fast and accurate lighting simulations.
A new computational framework simulates each stage of tree combustion, capturing detailed geometry and effects on branches and leaves. The method efficiently models the kinetic behavior of plants and their reaction to heat release in the combustion process.
Researchers at Tel-Aviv University and Facebook develop a computational technique to animate still images with realistic facial expressions and emotions. The method enables subjects in a photo to come to life, expressing various emotions through subtle variations in facial expressions.
A team of computer scientists has developed a novel, compact single-shot hyperspectral imaging method that captures images using a conventional DSLR camera equipped with an ordinary refractive prism. The new method achieves quality images without compromising accuracy, making hyperspectral imaging practical for ordinary users.
Researchers developed a computational framework that automatically suggests complementary parts and placement for 3D modeling, improving efficiency and user control. The method can learn from un-annotated online data, making the process more efficient.
A team of computer scientists developed a novel computational system to aid the design and fabrication of wind-up toys. The new method automates the construction of intricate mechanical assemblies, resulting in more compact mechanisms that consume less energy.
A Chinese research team used the Sunway TaihuLight to simulate a devastating 1976 Tangshan earthquake. The simulation created detailed 3D visualizations and informed engineering standards for buildings in seismic zones.
Shaden Smith and Yang You have been awarded prestigious fellowships for their work on high-performance computing applications. They are recognized for advancing the state of the art in sparse tensor factorization algorithms and scalable machine learning algorithms.
A pair of new computational methods developed by researchers from MIT, University of Toronto, and Adobe Research generate simulations that match real-world behaviors at rates 70-times faster than previously possible. These methods can automate the design process used to create dynamic mechanisms for controlled jumping.
A team of UC Berkeley researchers has developed a new technique for real-time user-guided colorization using deep neural networks. The system enables novice users to quickly produce reasonable results by learning common colors for different objects and making recommendations in real-time.
Ramesh Raskar has made significant impact in computational imaging, light transport and augmented reality with techniques like femto-photography. His work has led to applications such as EyeNetra, a low-cost technique to measure human refractive error and prescribe eyeglasses.
Bernd Bickel, an Assistant Professor at IST Austria, will receive the Significant New Researcher Award for his contributions to computer graphics and facial modeling. His research has introduced new techniques for capturing and reproducing facial structure and motion in 3D digital models.
Jessica Hodgins, a pioneer in character animation and robotics, has been recognized with the prestigious Steven Anson Coons Award for her outstanding contributions to computer graphics. Her work on physical simulation and control has advanced the development of realistic computer animations and human-robot interaction.
A global team of computer scientists and engineers developed an innovative technique for reconstructing complex objects using water displacement. The method accurately captures hidden parts of an object, previously inaccessible to laser scanners.
A new computational method called GRASS uses machine-learning techniques and artificial intelligence to automate the creation of plausible 3D shapes. The research paper showcases a generative recursive autoencoder for shape structures that can generate novel shapes without manual input.
A new tool allows users to design collapsible telescoping structures from any 3D shape, enabling compact mechanisms and deployable structures. The research, led by Carnegie Mellon Professors Stelian Coros and Keenan Crane, successfully prototyped flexible robot arms and tent-like structures.
A new image-based method reconstructs wiry objects in 3-D with higher resolution and accuracy, benefiting fields like animation, medicine, and topology extraction. The technique uses short connections to assemble wire configurations, exploiting unique characteristics of wiry objects.