Sony has invited the University of Georgia to submit proposals for the annual Sony Research Award Program (https://www.sony.com/research-award-program).
These awards provide funding for cutting-edge academic research and help build a collaborative relationship between faculty and Sony researchers. With awards up to $150,000 per year for each accepted proposal, both the Faculty Innovation Award and Focused Research Award create new opportunities for university faculties to engage in pioneering research that could drive new technologies, industries and the future.
Deadline: September 15, 2017 by 11:59pm PDT
Faculty Innovation Award
Award: Upper $100,000 USD
Period: One (1) year
Description: Global research and development at Sony enables us to foster innovative ideas, which could ultimately lead to future technology advancements and company growth. In order to speed up and expand the creation of new ideas, we would like to partner with universities. This partnership will help cultivate advanced concepts and fertilize our own research and development. Sony Faculty Innovation Award provides up to $100K in funds to conduct pioneering research in the areas listed below. Please include the single most relevant keyword to your submission.
General Research of Interests – Keywords
- Acoustic Metamaterials
- Active Vision
- ADAS / Autonomous Driving
- Adversarial Reinforcement Learning
- Automatic Parallelization on Multicore Systems
- Autonomous WLAN Network Optimization
- Cognitive Vision
- Computational Display
- Computational Photography
- Conversational Agent
- Distributed Machine Learning
- Edge Computing for IoT
- Electro-acoustic Transducer
- Full-Duplex Wireless
- Functional Image Sensing
- Hearing Aids
- Human Computer Interaction
- Human-in-the-loop Online Learning
- Immersive Audio Content Creation
- Knowledge-Based Information Extraction
- Motion Planning based on Machine Learning
- Multimodal Sensor Data Fusion
- Multitask Learning for Robotics
- Natural Language Processing
- Novel Quadratic Programming Algorithm
- Real-time Nonlinear Optimization
- Robust and Precise Force Sensor
- Scalable Neural Network Engine
- Small Proximity Sensor
- Speech Recognition
- Spoken Language Understanding
- Tactile Sensor with High Spatial Resolution
- Telepresence/Tele-existence System
- Transfer Learning and Domain Adaptation
- Ultra Low Power Processor for Edge Device
- Unlicensed Band (5GHz/2.4GHz) Co-existence
Focused Research Award
Award: Upper $150,000 USD
Period: One (1) year, with a possible extension
Description: Solid research is the underlying driving force to crystallize fearless creativity and innovation. While we are committed to run in-house research and engineering, we are also excited to collaborate with academic partners to facilitate exploration of new and promising research. The Sony Focused Research Award provides an opportunity for university faculty and Sony to conduct this type of collaborative, focused research. The award provides up to $150K in funds, and may be renewed for subsequent year(s). A list of candidate research topics appears below.
Adaptive Network Structure Optimization for Deep Learning
- Description: Deep Learning has already been used widely for a variety of practical applications. However, Deep Learning requires a lot of effort to optimize its network structure through trial and error in order to achieve the best results for each target application. The creation of a new and improved method is therefore being sought which adaptively optimizes the structure of deep neural networks, depending on the target training data. Sony is looking for innovative approaches beyond hyper parameter tuning in order to establish a new methodology.
- Scope of Proposal: Various approaches to predict an optimal deep neural network. The predicted network should be able to
- Solve a supervised discrimination problem with Deep Learning,
- Minimize the expected loss of the recognition rate for an un-known data set within a specific computational cost, and
- Be trained by various task sets (data, labels, network structures, and recognition rates).
Next-Generation Dialogue Management System
- Description: One of the ultimate goals of AI is the realization of natural dialogue between humans and machines. We recognize the advancement of technologies for simple request-and-response based dialogue, such as command-response, question-answering, etc. However, the product of these technologies to date is still far removed from natural communication with natural dialogue. Sony is looking for a next-generation dialogue management system which will be substantially more advanced than existing request-and-response based dialogue systems.
- Scope of Proposal: Research that is related to next-generation dialogue management systems which includes, but is not limited to
- General dialogue state description in goal-oriented dialogue system for multi-purpose applications; dialogue breakdown detection and recovery; conversation modeling for robust speech recognition, spoken language understanding, and/or dialogue state tracking; Q&A from unstructured dialogue history,
- Multi-modal dialogue system; in which dialogue state is modified by shared visual input e.g. dialogue while watching a calendar; which can generally “ground” users utterance to a robot’s complex behavior e.g. object of interest, speed of behavior, shape of trajectory, etc.
New Robotic Actuator
- Description: The actuator is one of the key components for robotic applications. The combination of an electro-magnetic motor, reduction gears (such as a harmonic drive), and an encoder is the most popular way to construct a traditional actuator, but it is still heavy and costly. To make robots ubiquitous, a new actuator that is safer and less costly is needed, where a new driving principle (other than electro-magnetic) is employed and the reduction mechanisms (if needed) are integrated. Sony is interested in new actuators including artificial muscles, variable-impedance actuators, ultrasonic motors (USMs), and other actuators based on novel concepts.
- Scope of Proposal
- To realize an actuator that is able to control position (angle), (angular) velocity, and force (torque)
- To realize an actuator whose power-to-weight ratio is greater than that for the combination of an electro-magnetic motor and a reduction gear
Machine Learning Based Image Processing
- Description: Recent advances in machine learning have created a paradigm shift on many applications. For instance, deep learning-based approaches have achieved a big leap forward over previous state-of-the-art technologies in image classification, image segmentation and image recognition. Sony is looking for innovative research in image processing based on machine learning to significantly improve existing image processing techniques and applications.
- Scope of Proposal: Topics of interests include image restoration, image transformation, image enhancement, noise reduction, color restoration, contrast enhancement, super-resolution, demosaicing, frame rate conversion, fusion of heterogeneous image sensors such as sensitivity improvement and color super-resolution, video compression, etc. with significant improved performance over existing technologies.
Next Generation Novel Display System
- Description: Nowadays, the degree of freedom to represent visual content is expanding in various directions. Fundamental visual parameters such as resolution, dynamic range, color gamut and frame rate are evolving up to 8K, HDR, BT.2020 and 240fps, while the recordable field of view (FoV) is now expanded to 360 degrees. Furthermore, the degree of freedom of viewpoint is also increasing by storing the light field or 3D modeling of the object. While the display system is also evolving with this advancing freedom for visual representation, there is still room to extract its potential. Sony believes it can create new customer value if a display system can be developed which enhances visual expression with practical cost and size.
- Scope of Proposal: Next generation novel display system which
- Provides immersive visual experience with enhanced quality of any kinds (depth, dynamic range, resolution or other unique functionalities) largely beyond the conventional display
- Can be potentially used in home/living room with multiple viewers
- Employs smart and unique computation to maximize quality
- Includes, but is not limited to, advanced Light Field/Plenoptic display, Holographic display (video), Screenless display and Image Sensor Integrated Display
- Excludes HMD or near-eye type
Image-based Photo-realistic Rendering
- Description: Sony is already successfully creating a realistic virtual world in the world of gaming with PlayStation 4. In the world of outside of gaming however, a challenge still exists to capture a three-dimensional space in the real world and reconstruct it for application in various fields such as movies, gaming and visual entertainment. Although some techniques are available to perform the rendering from 3D model and 2D image, they are not developed enough yet to perform a photo-realistic rendering that is indistinguishable from the real image in real time. This is due to the huge computational complexity. If such a photo-realistic rendering is developed with reasonable computational costs, it is feasible to reproduce the real world on the client devices and bring attractive, high quality features such as free viewpoint viewing, re-lighting, etc.
- Scope of Proposal: Photo-realistic rendering
- Image-based instead of artificial computer graphic. Computer graphics technique can be utilized for rendering
- Implementable on the latest PC/GPU
- Image quality equivalent to a 4K still image
Channel Modeling of Self-Interference Signals for In-Band Full-Duplex Systems
- Description: Sony is seeking essential wireless access technologies for beyond-5G mobile networks. Much attention has been paid to ‘in-band full-duplex’ to realize higher spectral efficiency and low-latency communications in next generation mobile networks. Deployments of the in-band full-duplex systems with interference cancellation and channel model assumptions will be important. However, practical channel models of self-interference signals have not been developed yet either in academia or in industry.
- Scope of Proposal: To investigate and develop channel models of self-interference signals for in-band full-duplex systems, taking into account the following aspects
- Large-scale (static) and small-scale (dynamic) characteristics, and
- Not only signals inside/nearby a device but also signals transmitted by a device, reflected by surrounding objects, and returned to the device itself.
Core Technology Design for Low Earth Orbit (LEO) based Satellite Communication
- Description: Sony is seeking new satellite communication technologies based on LEO which is located 1,200 km above the Earth’s surface. Satellite communication is currently considered to be one of the more important technologies for future wireless communication, much like 5G is. Since satellite communication can have an ultra-wide range and high communication capacity by broadcasting or multicasting, these significant benefits can be used to develop new mobile network applications for IoT (Internet of Things), UAVs (Unmanned Aerial Vehicles), and V2X (Vehicle to anything) communications.
- Scope of Proposal
- PHY/MAC layer technologies which have affinity with 4G/5G cellular system.
- Initial access procedure including random access, power control, timing advance, etc.
- To realize low power uplink communication in comparison with conventional solutions
- Interworking with the terrestrial networks i.e. 4G/5G
Millimeter Wave based Wireless Communication Technologies for Automated Vehicle Management
- Description: Sony is seeking wireless communication technologies based on the 5G cellular network for automated vehicles. Recently, the market demand for not only ADAS (Advanced Driver Assistance System) but also V2X (Vehicle to anything) communication has been increasing rapidly. In order to support an advanced driving environment in the future, new V2X applications (e.g. platooning, sensor data sharing, and remote driving, etc.) must be realized. The solutions should achieve the desired performance such as higher reliability, lower latency, and higher data rate in comparison with existing technologies such as DSRC (Dedicated Short Range Communications) and 4G cellular network based V2X.
- Scope of Proposal: Topics of interest include
- Define target use cases and their related problems to be solved with solutions to these problems
- Designing physical layer protocols and corresponding mechanisms
Egocentric Motion Tracking
- Description: Environmental sensor-based capturing systems are already available with very high accuracy and they detect the full-body motion of professional users (such as actors and players). On the other hand, there are increasing demands on wearable sensors which detect body motion precisely for a wide variety of applications such as research studies of sports training, rehabilitation, recording of action and labor, interaction for gaming, etc. We focus on cutting-edge sensor technologies and signal processing techniques and applications.
- Scope of Proposal: Motion tracking which
- can be applied to sports training, game entertainment, activity recognition, and computer interfaces,
- works robustly both indoors and outdoors (environment-independent),
- may employ multi-modal sensing (IMU, camera, magnetometer, barometer, proximity, etc.),
- may employ multiple sensors (but must be lightly-equipped),
- is robust to fast motion, self-occlusion, and twist motions, and
- may be combined with force sensing/feedback and haptic applications
3D Generic Real World Object Recognition
- Description: Vision-based object recognition is one of the key technologies for Augmented Reality (AR) which can provide a variety of useful information to users in support of daily activities. However, current object recognition technologies, which are primarily based on 2D vision recognition algorithms, have limitations in detecting the 3D shape of objects including curved surfaces. We ideally seek a technology that can recognize a real-life 3D generic object as it appears in a wide variety of scenes in order to display more accurate and advanced information to users. Sony is looking for various approaches and/or technologies to realize this challenging application: 3D generic real world object recognition.
- Scope of Proposal: Vision-based 3D object recognition algorithm which
- detects 3D shape of real world objects as they are without markers,
- can segment, identify, and classify objects associating them with the name of the object,
- works equally well with either indoor or outdoor scenes,
- has low latency with tracking capability at 60 frames per second (fps) or higher,
- has robustness in occlusions, busy scenes, size of objet, and shape changes, and
- can create scalable dictionaries automatically.
High-Precision Sensing Technology of Human Body Motion in Living Spaces
- Description: There are systems that detect motion of the human body and provide functions adaptively, but the type of motion detected by ordinary image recognition is limited so it is difficult to realize a system that understands a user’s conscious or subconscious behavior. If we can sense the motion of the human body with high accuracy in the living space continuously and constantly, we will be able to realize a system that comfortably carries out advanced support for all of life’s activities.
- Scope of Proposal
- High precision human body motion detection system
- High spatial and temporal resolution (can recognize detail gestures)
- Available in the user’s living space (no use of markers, etc.)
- Biologically safe, has low power consumption, is easy to install
- User identification function
AR/MR Eyewear Display
- Description: AR/MR (augmented reality/mixed reality) eyewear technology is evolving rapidly. One of the big challenges to achieve high viewing reality and wearing comfort is to realize a compact AR/MR see-through eyewear display with a wide field of view. It is well-known that there are trade-offs between the size of AR/MR optics and the field of view provided as well as between optical efficiency and see-through transmittance. Sony is therefore looking for innovative approaches which incorporate a wide field of view, high optical efficiency, clear see-through capability, and a compact display size.
- Scope of Proposal: Full color AR/MR eyewear display with a wide field of view which potentially has
- High see-through transmittance,
- High optical efficiency, and
- Compact form factor for digital eyewear products.
- Eligible Applicants: We accept applications from Principal Investigators (PIs) who are tenured or tenure-track faculty members and are eligible to supervise Ph.D. students at a university/educational institution in the USA or Canada. Full professors, associate professors, and assistant professors are eligible to apply.
- Co-PI(s): We accept applications that have a PI and one or more co-PIs for the same proposal. However, only one award is made to the primary PI and the primary PI’s university/institution if the proposal is selected. All Co-PIs must meet the same eligibility criteria as that for a PI, and Co-PIs will be required to sign program documents.
- Multiple Proposal Submissions: A PI or different PIs from the same university/institution may submit more than one proposal for different research topic(s). However, please do not submit identical proposals more than once.
Please include the required items listed below in your submission:
- Title, abstract, methods, goals/milestones, references, and either one Focused Research Theme (if for the Focused Research Award) or one primary keyword (if for the Faculty Innovation Award).
- Please describe differentiation from the current state-of-the-art
- Please include the best contact email address and phone number for the PI
- The CV for the PI must be included when you submit a proposal. The CV file must be a separate file from the proposal file. There is no page limitation for the CV, however the CV file size must be under 16 MB.
All proposal contents (not including the CV) must fit within 11 pages (a ten-page maximum proposal with references and a one-page budget summary). The file format must be a PDF or a MS Word file format and must be under 16 MB in size.
- Submission must be done through the online submission form (click to open).
- Submissions by email will not be accepted.
- Submitters will automatically receive a confirmation email once they complete the online submission form and process. Please keep the confirmation email as proof of each submission.
Please read all Guidelines and Requirements before preparing your proposal.
Please send an email to the Research Award Program Administration Office at ResearchAwardProgram2017@am.sony.com if you have a question regarding the 2017 Sony Research Award Program. The Research Award Program Administration Office is the only resource that can officially answer your question(s).