July 8, 2015 July 9, 2015

Faculty Summit 2015

Location: Redmond, WA, USA

Thursday, July 9 | 9:45 AM-1:00 PM

  • Accurately predicting body fat, waist size, and BMI

    Body measurements are important for health assessments and weight control. This project accurately obtains body measurements without measuring devices. Participants match their body type to a computer-generated image, and the system deduces their waist size and Body Mass Index (BMI).

    Beyond A/B tests: multiworld testing as a service

    We demonstrate a “decision service” built on multiworld testing, which can answer much more detailed questions than A/B testing can. Designed to interoperate with existing systems and scenarios, the service includes multiple consumable pieces and a standalone cloud service.

    Deep models for automated melanoma diagnosis

    Early detection of melanomas is crucial to clinical outcomes. Using deep probabilistic models trained on images of moles, our system distinguishes benign lesions from malignant tumors, surpassing the diagnostic ability of previous computational attempts and experienced dermatologists.

    Feature discovery and exploration for interactive ML

    Our platform promotes quick interaction between human teacher and learning machine. It uses an interactive loop, where users alternate between feature ideation to fix errors in the training set and exploration to discover new pockets of false positives and false negatives.

    Fully articulated hand tracking

    We present a real-time, articulated hand tracker that enables new possibilities for human-computer interaction. Our system accurately reconstructs complex hand poses, using only a single depth camera. It also allows for a high-degree of robustness, continually recovering from tracking failures.

    Image captioning with near-human performance

    People can look at a picture and immediately describe what is important. This project gives computers similar capabilities, combining the latest in deep neural network image processing with the latest approaches to language modeling to produce an automated captioning system.

    SysSieve: turn user feedback to actionable insights

    Microsoft customer support specialists handle issues captured by free-form and unstructured text input. SysSieve employs machine learning, natural language processing, information theory and ontology modeling to derive automated inference and insights from a range of free-form text.

  • Crypto-verified elections: reality in Texas

    Learn how the STAR-Vote system will soon enable voters in Austin, Texas, to verify the accuracy of election results directly, without relying on election officials and system vendors. Oh, and it will save them money, too.

    Eye Gaze wheelchair

    Begun during the Microsoft Summer 2014 Hackathon, Eye Gaze technology strives to bring independent mobility to people with disabilities who cannot use a joystick.

    Minecraft in education

    Learn how Minecraft is being used in education. From math and science to humanities and history, Minecraft is empowering teachers and engaging students to create, collaborate and share.

    Project Premonition: detecting emerging infectious diseases

    Project Premonition seeks to detect pathogens before they make people sick. It uses drones and robotic traps to capture mosquitoes, which are potent disease vectors, and then detects pathogens by gene sequencing the collected mosquitoes and computationally searching for pathogens.

    qCards: low-cost audience polling using computer vision

    We present a low-cost alternative to electronic response systems such as i>clicker. Our system uses paper cards for voting and a low-cost smartphone or a laptop plus webcam to capture and process votes.

  • Be there: 3D audio virtual presence

    Using 3D scans of your head, we synthesize your personal head-related transfer function. Then you don headphones and stand before of a pair of large screens, experiencing sights and 3D sounds as if you were there.

    Multi-touch interaction using dynamic chord gestures

    DynaChord allows the use of dynamic, hierarchical chord gestures to interact with a large multi-touch display. We show how DynaChord’s in-context visual feedback and feedforward help the user learn the gestures and actively assist in their performance.

    Presenter Camera: a virtual camera for best presentation

    See how our Presenter Camera addresses the bane of presentation videos: either you see the presenter but not the digital contents, or you watch the stream of the digital contents, but you don’t know what the presenter refers to.

    Rapid prototyping with TouchDevelop

    Devices are everywhere: from those we use, like phones and tablets, to those we can program, including Arduinos and IoT sensors. TouchDevelop now supports the creation of device firmware—so even beginners can use devices to program devices.

    RoomAlive toolkit

    The RoomAlive Toolkit is an open-source software development kit that enables developers to create new immersive augmented reality experiences by calibrating a network of multiple Kinect sensors and video projectors.

    Secure, anonymous and instant communication between phones and touch

    We present a new technique for pairing devices using the phone’s camera and capacitive touch sensing. Our technique provides rapid, secure pairing with device tracking and targeted interactions, enabling a range of interactive functions.

    Textile electronic touch sensor

    We show designs that deliver multi-touch performance from a textile platform and demonstrate how these touch sensors offer unique functionality to next-generation electronic devices.

  • Artificial intelligence at Microsoft

    Learn about artificial intelligence research, applications (e.g., Hyperlapse) and datasets (e.g., MS COCO) that you can engage with, and see their impact in such Microsoft products as Bing, Cortana, Kinect, HoloLens, and Oxford.

    Azure for Research and Azure Machine Learning

    Amp up your research. Microsoft Azure for Research offers grants that enable big data computations in the cloud. Microsoft Azure Machine Learning makes it easy for the academic community to use machine learning technology for their research.

    Data Science at Microsoft Research

    Microsoft has products, tools and services that promote data science research, education and insights, and we’re eager to facilitate collaborations with the academic community by supporting technology programs and data. Learn what’s on the horizon in data science.

    Open Source Software from Microsoft for Academics

    Open source is a powerful way of advancing software development. Microsoft has open sourced cutting-edge research projects as well as such key software as .NET, and has classified it especially for academics. Often the software is cross platform, too.

    Project Catapult: FPGA-based enhancement of server hardware

    Project Catapult is a ground-breaking FPGA-based enhancement of server hardware that vastly improves the performance and speed of datacenter workloads. We will demonstrate its power by running workloads on both a standard server and a server enhanced with Catapult.

    Cognitive Services

    Cognitive Services offers an evolving portfolio of APIs and SDKs, enabling developers to add intelligent services to their solutions. The current portfolio includes services for facial recognition, speech, computer vision and natural language.

    Research Fellowships

    The Microsoft Research Fellowship Program provides financial support for students and early-career faculty, allowing them to focus on their research. Fellowships build lasting relationships between Microsoft Research and academic institutions.

    Research Internships

    Interns at Microsoft Research put theory into practice, while working with renowned researchers and networking with fellow PhD students from around the world. Microsoft Research Internship opportunities are available year-round at all our worldwide labs.

    Academic Services

    Academic Services is a suite of apps and services that will help researchers and students be more productive and stay current with their research domain and related areas.

    What’s it like to work at Microsoft Research?

    Learn about internship, fellowship and career opportunities that put you on the forefront of computer science research and technology development, let you see your work impact real-world applications and enable you to collaborate and network with researchers around the globe.

  • Project Flashprog: data transformation by example

    Project Flashprog develops technologies for extracting data from semi-structured documents into a relational table, which can be used for analysis or visualization tasks. An intuitive interface allows users to express their intent, explore the resulting program and tailor it to their needs.

    Microsoft Research Computational Photography applications

    The Microsoft Research Computational Photography Group uses computation to push the limits of photography, conducting early-stage research and turning promising research into applications. The demo will showcase some of our recently shipped applications and walk through the technology behind them.

    NUIGraph: explore and display your data with touch

    Big data offers immense potential for insights, but it can be challenging to interact with and analyze. By combining visualization and analytics, NUIGraph helps uncover answers even when we don’t know exactly what questions we could be asking.

    Power Map

    Power Map provides robust analytical tools for big data. From the familiar base of Excel, you can map data, discover new insights, and then create and share stories about that data.

    Prediction Lab

    An interactive platform that combines data from a variety of sources, Microsoft Prediction Lab enables people to produce and consume information about the present status and predicted outcomes of a wide range of market intelligence questions.

    Microsoft Band: Reimagining personal healthcare

    Built on years of research, testing, iterating and inventing, Microsoft Band represents an entirely new product category for the company. Using a security-enhanced cloud platform, Microsoft Band stores and analyzes personal health data, potentially transforming personal health and fitness. See how to use Microsoft Band SDK to write applications that communicate with the Band.

    Removing the language barrier with Skype Translator

    Incorporating the latest advances in machine learning, speech recognition and automatic translation, Skype Translator seeks to remove the last hurdle to global communication: the language barrier. See it in action and learn how it works.

  • Code Hunt: exploring how programmers code

    Code Hunt is an education game whose 140,000 participants have generated over 1.5M programs, which we can link to specific levels of expertise. Discover how coders code and how technology can make the process more accurate and less painful.

    Madoko: create and share programmable documents

    Madoko lets you write papers in advanced markdown and get both beautiful HTML and PDF output. It promotes collaborative editing, with concurrent changes automatically shown and merged. Madoko also runs on the web and automatically synchronizes changes in the cloud.

    Node Atlas

    Node Atlas is an agile, flexible geographical Azure database with a web front-end that allows users to curate and explore countless pieces of information indexed by geography, date, time, and any number of metadata items.

    Programmability of scalable, geo-distributed, interactive applications

    Orleans enables programmers to build scalable, interactive applications that run across hundreds of servers in a datacenter. We show new mechanisms that enable Orleans applications to scale across many datacenters, with improved fault tolerance and latency at no cost to the programmer.

    Spatial search optimization

    Query from Cortana is often expanded into a complex spatial search reflecting both geographic intent of current location and geo-specific keyword rewrites. See how a cost-based optimization vastly outperforms existing techniques motivates predictive optimization deployed to Bing from our prior work.

    Time-travel debugging for JavaScript

    Our Time-Travelling Debugger for client-side HTML/JavaScript applications lets developers step forward and backwards through time. It leverages memory management and virtualization features so that developers enjoy a much faster cycle for bug hypothesis generation and investigation, with negligible runtime overhead.

    Trill & Tempe: high speed temporal analytics and visualization in real time

    We demonstrate how Halo debugged their multiplayer application using two Microsoft Research technologies: Trill, a streaming query processor, and Tempe, a Web-based interactive, real-time dashboarding and analysis tool.

    Verifying networks like programs

    Network configuration errors frequently cause outages. Our tool, Batfish, detects a broad range of such errors proactively, before the configuration is applied to the network. Learn how Batfish analyzed bugs in two large university networks, finding many misconfigurations.

    The design and implementation of a wireless surveillance-based system

    In smart cities, cameras will generate copious amount of video data, but their ubiquity limits video that can be sent to the cloud, especially on wireless networks. See how a real-time distributed wireless surveillance system uses edge computing to save wireless capacity.

    What can we solve with a quantum computer?

    Wondering what quantum computing actually requires and the high-level physics behind it? This demo explains the software architecture and shows how to run small quantum algorithms on your laptop through simulation.