Projects

The Activity Detector is a computer vision system that extracts and analyzes truck motion segments in real-time from continuous video feeds at facility gates. It uses technologies such as OpenCV, YOLOX, Triton Server, and Kalman Filter for image processing, object detection, and motion tracking. This Python-based solution utilizes PyTorch for deep learning and PostgreSQL for data management.

To achieve high detection accuracy, the system incorporates visual similarity and deep matching techniques, enabling reliable truck identification and tracking in various conditions. Implemented in production environments, it operates via Kubernetes on the Amazon cloud platform, allowing for scalability and reliability across different operational scales.

Construction progress monitoring requires aligning 3D scans with design models, but manual alignment is slow and error-prone. Our innovative system automates this process using fiducial markers. We place circular markers on surveyed control points and rectangular markers on walls. Our software detects these markers in 2D projections of 3D scans, then calculates the transformation between scan and model coordinates. This approach leverages existing survey points and uses error-correcting codes for reliable detection in challenging environments. Our system completes alignments in minutes instead of hours, enabling near real-time progress monitoring and deviation analysis. By automating this process, we dramatically accelerate quality control and progress tracking, bringing real-time analytics to construction sites. This solution helps construction teams quickly compare as-built conditions to design models, revolutionizing project monitoring and management.

This project involves the development of an innovative TV remote control featuring a clickable touchpad surrounded by ridges. The design evolved through multiple iterations, incorporating user feedback and technological advancements. Key features include improved navigation through physical affordances, a handwriting-based channel input system to replace traditional number buttons, and haptic feedback for enhanced user experience. The project spans from initial concept to production, including extensive research, prototyping, and user studies, resulting in multiple publications and patents in Smart TV remote technology.

DRIVE is a novel interaction method, which allows users to reach behind a display and manipulate virtual objects with their bare hands. The system combines advanced technologies, including anaglyphic stereoscopic rendering, face tracking, and hand tracking, to create a seamless and intuitive user experience. The prototype system uses a transparent LCD panel, a depth camera, and a physics engine to enable users to interact with both real and virtual objects. The system was demonstrated at the Samsung Tech Fair 2010 and has been published in the IEEE Symposium on 3D User Interfaces and the IAPR Conference on Machine Vision Applications.

Flatworld is a wide area mixed reality system that uses large rear-projection screens to create immersive virtual environments. Users can freely walk and run through simulated rooms, buildings, and streets, interacting with both physical props and digital graphics. The system can simulate a variety of locations and situations, and it includes immersive audio, 4D sensory features, and real-time graphics synchronization. Flatworld has been used for training purposes, and its development has been documented in several publications and conference presentations. The system has also been recognized with awards for its innovative approach to immersive visualization.

Virtual Iraq is a virtual reality exposure therapy application for treating PTSD in Iraq war veterans. Built on the FlatWorld Simulation Control Architecture (FSCA), it creates realistic, customizable Iraqi environments. The system leverages FlatWorld's innovative integration of physical props with digital displays, allowing clinicians to control various stimuli for personalized therapy. Utilizing FSCA's flexible, scalable approach, Virtual Iraq provides an immersive, effective treatment tool that has shown promising results in reducing PTSD symptoms among veterans.