Researchers developed neural networks to evaluate short narratives, improving predictions over a baseline system. The AIs classified texts into popular and non-popular categories, highlighting the importance of understanding story structures in narrative evaluation.
Researchers at Disney Research developed an algorithm that can automatically recognize soccer formations and defensive strategies from player tracking data. The algorithm outperformed conventional methods in identifying dynamic player roles and coordinated team behavior, with applications beyond sports.
Machine learning scientists at Disney Research developed a dynamic word embeddings model that uncovers how the meanings of words change over time. The model, which integrates neural networks and statistics used in rocket control systems, detects semantic change throughout history by analyzing semantic vector spaces.
A new tool from Disney Research enables novices and experts alike to create high-quality motion cycles in a matter of minutes, enhancing the creative process and expanding artist contributions. The tool uses an algorithm to extract the motion cycle from a performance, allowing for editing and customization.
Disney Research's method enables artists to design articulated, physical versions of animated characters with desired motions and poses. The team demonstrated their approach by designing a 2D puppet-like character and a gripper that can pick up light objects.
A computational design tool developed by Disney Research enables the creation of compliant mechanisms through flexibility, offering potential advantages in industry. The tool optimizes device performance while ensuring robustness and reducing strains.
The Magic Bench enables users to see, hear, and feel animated characters in a shared 3D space. Researchers have created a platform that combines augmented and mixed reality for a seamless experience.
A new technique called factorized variational autoencoders (FVAEs) uses neural networks to predict a viewer's facial expressions for the remainder of the movie after observing an audience member for just a few minutes. The approach demonstrated impressive accuracy in predicting reactions, outperforming conventional methods.
Researchers at Disney Research and UC Davis have developed a method for computer vision programs to understand spatial relationships in images based on caption sentence structure. This approach enables accurate visual localizations for language inputs, outperforming baseline systems that do not consider natural language structure.
Researchers developed a new AI-based denoising model to eliminate noise in computer-generated images, enabling production-quality rendering at faster speeds. The system was trained on millions of examples from the film Finding Dory and successfully removed noise on test images from different films.
Researchers found that pre-school children respond well to simple interactions, while older children engage with robots that reference previous conversations. Interactive storytelling also boosts vocabulary and cognitive development in young children, suggesting that collaborative technology can be an effective tool for education.
The research enables immersive viewing of pre-rendered video in six degrees of freedom using a head-mounted display, making it easier to repurpose animated film assets. The novel video format can be rendered in real-time from an arbitrary point of view, allowing for motion parallax and interactive experiences.
Researchers at Disney Research have demonstrated an ultra-wideband backscatter communication system that enables IoT sensors to transmit data via ambient radio waves. This approach radically reduces the power requirements of sensor nodes, allowing for widespread deployment in metropolitan areas.
A new innovative method developed by Disney Research makes it possible to realistically simulate hair by observing real hair in motion. The framework considers the dynamics of hair and can be used with a wide variety of simulation methods.
The Disney Research team has developed a new system called Makeup Lamps that can track an actor's facial movements and apply light-based makeup in real-time. This technology enables the creation of stunning transformations without physical makeup, with potential applications in theater, film, and other creative fields.
Disney Research scientists developed a method to catch a real flying ball using virtual reality, demonstrating enhanced user experience. Different visualization styles were tested, revealing that users caught balls in all three modes with varying strategies.
A new automated method based on deep learning techniques analyzes detailed game data to create models of how a typical player would behave in a given situation. This allows for the comparison of actual player behavior with predicted ghostly behavior, providing valuable insights into defensive athletic performance.
Researchers developed a method called quasistatic cavity resonance (QSCR) that enables wireless power transmission throughout a room. This technology can charge multiple devices simultaneously, including smartphones, fans, and lights, using near-field standing magnetic waves.
A new technique enriches animations by simulating fine detail and smooth large-scale appearance of granular materials. The method, developed by Disney Research, ETH Zurich and Dartmouth College, significantly expands the number of grain types that can be rendered together.
A Disney Research team has developed a model-based method to realistically reconstruct teeth for digital actors and medical applications, even with obscured teeth in photos or videos. The new method uses statistical modeling to create natural variations in tooth shape and spacing.
Disney Research's CANVAS and Story World Builder tools help authors synchronize characters, fill plot holes, and create virtual worlds. These graphical platforms simplify the creation of story worlds, making it possible for anyone to tell an animated story.
Researchers at Disney Research developed an AI-based system that can automatically learn the association between images and sounds, with applications in film sound effects and aiding visually impaired individuals. The system uses video data to filter out uncorrelated sounds and learns which sounds are associated with an image.
A novel color unmixing algorithm cuts time and boosts quality in film compositing by extracting multiple color layers. The method requires only one-tenth of the manual editing time of professional tools, offering superior results.
A Disney Research team created an automated method of crowdsourcing multiple lines of dialogue for robots, using a persistent interactive personality that can translate high-level goals into simple narratives. The method was tested in a trivia contest and resulted in 716 unique lines of dialogue.
Disney Research has developed a VR360 player with enhanced haptic feedback, allowing users to customize and personalize sensations. The application enables full-body sensations and a wide range of 'feel effects' that can be triggered by user movements or biofeedback.
Researchers at Disney Research have found a way to reduce the time and effort required to train facial performance capture systems. By using synthetic data generated from a small sample of recordings, they were able to determine a set of training data that was tens to hundreds of times smaller than normal without affecting accuracy. Th...
Disney Research has developed a one-legged hopper that runs on battery power, breaking its dependence on off-board power. The robot weighs less than five pounds and can maintain its balance for approximately seven seconds.
Disney researchers developed a speech technology system that can sort through overlapping speech, social side talk, and creative pronunciations of young children to make it work. The system was 85% accurate in recognizing keywords, outperforming commercial speech recognition systems.
Researchers developed a perceptual model to predict the perceived softness and stiffness of nonlinear elastic objects, replicating an object's feel despite material differences. The model was validated through experiments and shown to accurately predict how people perceive the softness of various materials.
A new technique developed by Disney Research can capture the crucial details of the eyes with just a single facial scan, producing highly realistic eye reconstructions. The parametric model enables flexible manipulation of the captured data to simulate emotions and fatigue.
Disney Research's new adaptive polynomial rendering method improves Monte Carlo ray tracing by varying polynomial functions based on image complexity, reducing noise and preserving fine detail.
A single camera can capture high-quality facial performance using a new method developed by Disney Research, which takes into account the underlying facial anatomy and skin thickness. This approach allows for robust and accurate tracking of facial expressions without requiring multiple cameras or extensive pre-computed facial motions.
The new compiler system allows users to program industrial knitting machines with simple shapes and translate them into needle-level instructions. This enables the creation of a variety of 3D knitted shapes, previously difficult to produce due to complex machine control.
A new modular design approach simplifies the creation of acoustic filters, allowing for rapid prediction of acoustic properties in complex assemblies. This enables the development of innovative applications such as audio tags and wind instruments with customizable sound profiles.
A new character animation technique has been developed by Disney Research, eliminating unsightly artifacts that occur when bending joints in computer animations. The method uses pre-computed centers of rotation to calculate skin deformation, minimizing volume losses and bulging.
A computational design tool developed by Disney Research enables users to create delicate, yet structurally sound, 3D objects with interconnected shapes. The tool helps identify and strengthen weak areas in the designs, making it possible for the general public to create beautiful artifacts that also withstand handling and shipping.
Researchers have demonstrated that consumer-grade LED bulbs can be modified to serve as both light sources and receivers of signals, enabling the creation of a network of devices in a room. This technology has potential benefits for interconnecting IoT devices without threatening the scarce radio spectrum.
Computer vision systems can now learn to recognize objects they have never seen before by analyzing word use and contextualization, reducing the need for thousands of labeled images. This new learning paradigm, called semi-supervised vocabulary-informed learning, was developed by Disney Researchers using a large dataset of English words.
Researchers developed a machine learning program that can detect human activity in videos sooner and more accurately than previous methods. The program is trained to gain confidence in its predictions over time, allowing it to detect activities earlier and with higher accuracy.
Researchers from Disney Research and Fudan University developed a new approach that enables computers to recognize events in videos, including categories of events they've never seen before. The software associates visual elements with each type of event and can learn from new examples to improve its accuracy.
Automated camera system improves sports broadcasts by learning from human operators and achieving smoother footage without jerkiness. The system uses a new approach called imitation learning, which repeats multiple times and analyzes deviations to learn from human mistakes.
A new type of hydrostatic transmission combines hydraulic and pneumatic lines to drive robot arms with extreme precision, enabling delicate tasks like picking up an egg. This technology enables robots to interact with people in a life-like manner, making them ideal for human-robot interaction research.
Researchers developed a software tool called DefSense that enables non-experts to create deformable objects with deformation-sensing functionality. The tool optimizes sensor layouts based on user-defined deformations, allowing for customized input devices.
Scientists at Disney Research have developed a method to differentiate between even seemingly identical electronic devices using their unique radio frequency emissions. The EM-ID system achieved 95% accuracy in identifying individual devices, with limitations including the need for powered-on devices and potential signal overlap.
Researchers at Disney Research and Carnegie Mellon University have developed a system to debug intermittent systems, such as those that harvest energy from the environment. The Energy-interference-free Debugger (EDB) can monitor and debug these systems without interfering with their power state.
Researchers developed Chalkboarding, a graphical user interface that enables users to quickly retrieve basketball plays from a database by sketching what they seek. The approach outperformed keyword-based searches in user testing, offering a more efficient and effective way to access sports analytics data.
Disney researchers create a low-cost, handheld depth sensor system for capturing 3-D models of marine life with millimeter accuracy. The system corrects for refraction effects and supports coral reef restoration efforts in the Bahamas.
Scientists at Disney Research developed an automated approach to generating life-like character motions, easing game designers' workload and providing instant feedback. The system seamlessly integrates into existing pipelines to produce virtual characters that can autonomously reason about their environments.
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 team from Disney Research and ETH Zurich has developed a system that can transform existing stereoscopic content into multiview content in real-time, enabling the use of glasses-less 3-D displays. The system uses image domain warping to synthesize eight new views in high definition.