The "Four Eyes" Lab, directed by Matthew Turk and Tobias Höllerer, pursues research in the four I's of Imaging, Interaction, and Innovative Interfaces. During the open house, we will be describing and demonstrating several ongoing research projects. Feel free to drop by any time during from 2:00pm to 5:00pm and check out any projects that might interest you.

List of Presented Projects:

Placement of Object Annotations in Video-Based Augmented Reality

Vineet Thanedar

Augmented reality involves supplementing our view of the real world with virtual or computer-generated information. In numerous applications, annotating objects in the scene can be of immense help. It aids our understanding of the scene in view. However, inefficient, sub-optimal placement of these annotations in the scene can lead to problems, such as occlusion of other important objects, visual clutter, or loss of annotation coherency. In this project, we attempt to solve the problem of  placement of annotations in a video-based augmented reality setup so as to avoid the problems mentioned above.  

A Non-Photorealistic Camera: Detecting Silhouettes with Multi-Flash

Rogerio Feris, Jingyi Yu (MIT), Adrian Ilie (UNC)

We describe a non-photorealistic camera ('NPR camera') that robustly captures shape edge information in real scenes. A fixed camera captures the scene under one flash at a time and we exploit the resulting shadows to classify the scene edges. We synthesize new stylized images by highlighting or de-emphasizing the detected edges.  

Probabilistic Expression Recognition on Manifolds

Ya Chang, Changbo Hu

We present a probabilistic video-to-video facial expression recognition based on manifold of facial expression. The manifold of facial expression is embedded in high dimensional image space. In the embedded space, images with different expressions can be clustered and classified by the probabilistic model learned on the manifold of expression. 

Campus Modeling and Visualization

Lihua Lin

We construct a 3D campus model for information visualization by integrating the Maya 3D models (or other 3D models which can be converted into VRML model) with CAD drawings. Elevation information is used to model the actual ground.    

Hand Gesture Input for Battuta

Mathias Kolsch

Project Battuta is an interdisciplinary research initiative to investigate the potential of emerging technologies and geospatial information resources to bring new functionalities to mobile field data collection.   At the 4eyes lab, we research novel user interfaces for wearable computing environments, in particular vision-based hand gesture input. A head-mounted display (HMD) provides a screen for data output, and a head-mounted camera (HMC) allows for data input via hand gestures, performed in front of the body of the HMC's wearer.  

Postural Comfort

Mathias Kolsch

While human factors research gives advice on the range in which humans can  operate without experiencing musculoskeletal strain, fatigue or  discomfort, no objective measure for "comfort" is known. In this project,  we defined a measurable foundation for comfort, which is usually simply  defined as the absence of discomfort. Our work suggests that  human-computer interfaces should be designed within the limits of a  comfort zone. Otherwise, using the interface can prompt the adoption of  alternative use patterns, which are often less favorable because they  trade off the unnoticeable potential of injury for comfort. We also  conducted a user study on the range in which humans prefer to operate  their hands when carrying out free-hand gestures. This study also served  as an example for how to design studies for comfort evaluation.

ARWin

Stephen DiVerdi, Daniel Nurmi, Comron Sattari

The goal of the ARWin project is to develop an augmented  reality office desktop environment.  The basic ARWin system allows for execution of traditionally desktop applications in an augmented reality environment, and provides a testbed for exploring novel application and user interfaces that take advantage of the extended 3D workspace. 

Hidden Infrastructure Visualization and Interaction Techniques in Outdoor AR

Ryan Bane

The objective of the research will be to produce an outdoor AR system to allow a user to visualize the occluded internal structure of a building (rooms, corridors, stairways, etc) from a location inside or outside and nearby the building. The user should be tracked and the view should be updated to reflect the user's position in real time. Due to the complex information displayed, techniques for interacting with the visualization will be explored. These techniques will include tools to allow the user to focus on information relevant to specific areas of the building, to explore depth relationships between visualized elements, and to mark up the visualization.

Face Detection, Alignment and Tracking for Real-Time Visual Interaction

Changbo Hu, Rogerio Feris

We describe a real-time system for detecting, aligning and tracking human faces. This is an attempt to make machines aware of people in order to improve human computer interfaces.

Face and Hand Tracking with Skin Color Segmentation

Haiying Guan

In this project, based on the results of skin color segmentation, a person’s face and hands are detected and tracked by the clustering technique. The user can control a visualization of a cylinder (size, shape, rotation) by movements of the face and hands. The project is also a partial work for a collaborative project with Media Arts and Technology: Interface Device for Interactive Initialization.

Temporal Integration for Continuous Multimodal Biometrics

Alphan Altinok

Once authenticated, should a user have universal access to resources for the complete session? What if an unauthorized user takes over the session? We attempt to achieve continuity in biometrics decision making by temporal integration based on propagating uncertainty.

Ryan Garver

This system uses a stereo camera and the TLib image processing  library to track a users hands. With this tracking, mouse pointing and clicking is possible. Future plans include a HMM based open arm gesture recognition system working along side with the mouse pointing capabilities.