July 18, 2016
QoS Enhancement of Wireless Video Networking
This project aims to further improve the quality of service (QoS) techniques for WiMAX/LTE 4G protocol stacks. We are working on a scheduling algorithm based on multiple-input multiple-output (MIMO) for not only transmitting traditional videos, but also transmitting videos with scalable coding. Moreover, the feedback resource allocation algorithm can be extended to a MIMO system…
Language Grounding by Understanding and Generating Narratives for Dynamic Environments
Language is given meaning through its correspondence with a world representation. This correspondence can be at multiple levels of granularity or resolutions. In this project, we study multi-resolution language grounding in multiple domains of sport commentaries, geometry questions and images. This project aims at building a framework to learn to understand and generate narratives for…
Diagram Interpretation and Reasoning via Spoon Feed Learning
Diagram interpretation, an essential element in question answering, is the problem of identifying visual entities, properties, relations and correspondences to specialized knowledge repositories. Previous work uses hand-engineered rules for understanding specific diagrams. In this project, we introduce a unified, end-to-end framework to diagram interpretation and reasoning that is applicable to a wide range of diagrams.
TransPhorm: Improving Access to Multi-Lingual Health Information through Machine Translation
The TransPhorm project is aimed at facilitating the production of multilingual health and safety information materials for individuals with limited English proficiency. We have developed human-computer collaborative translation management systems for public health workflow, domain adaptation methods for machine translation (MT) models and new quantitative frameworks for studying user preferences in MT.
Neuromorphic Computing
Neuronal networks are capable of processing specific data and tasks optimally and in real time. Many of these problems are computationally expensive to solve with current computing systems. We are developing algorithms and architectures inspired from the design principles of neurobiological networks to solve these problems more efficiently.
A Street-View Video Dataset
For research in autonomous cars and robotic applications, a street view video dataset with dense semantic labels will be very useful. Semantic labeling is vital for training models for object recognition, semantic segmentation or scene understanding. We developed a large-scale street-view suburban video dataset comprising over 400k images, and a 3D-to-2D label transfer method to…
Real-Time Gaze Estimation with a Kinect Camera
We developed a 3D gaze estimation system that can estimate the gaze direction using a Kinect camera. A 3D geometric eye model is constructed to determine the gaze direction. Experimental results indicate that the system can achieve an accuracy of 1.4~2.7 degree and is robust against large head movements. Two real-time applications (playing chess and…
3D Object Modeling Using a Kinect Camera
We investigated the use of Kinect cameras for acquiring images of an object from multiple viewpoints and building a 3D model of the object. We developed methods to take care of the problems of noisy depth data, symmetrical objects and objects with relatively few views. Simulation results show that accurate 3D models could be constructed.
3D Human Pose Estimation Based on Structure from Motion
This research aims to estimate 3D human pose by utilizing Structure from Motion (SfM), 3D modeling and pose estimation techniques. Structure from Motion is the process of estimating 3D structures from 2D image sequences or videos. In our approach, the camera produces multiple images by shooting video of the target while simultaneously reconstructing both the…
Activity Recognition
We propose a framework, consisting of several algorithms, to recognize human activities that involve manipulating objects using a Kinect camera. Our algorithm identifies objects being manipulated and models high-level tasks being performed accordingly. We evaluate our approach on a challenging dataset of 12 kitchen tasks that involve 24 objects performed by 2 subjects. The entire…
Previous page Next page