USC Computer Vision -- Major Research Topics and Projects

Current Research Areas or Projects on Top, Older Ones on the Bottom

Context Tracking Visual tracking in unconstrained environments is very challenging, exploit context with Distracters and Supporters. (Recent Results) (Software Release)
Model Based Tracking Focus on Pedestrians and Vehicles with Part-based Detection. Motion regions do not always correspond to objects (people or vehicles). Single objects may split into multiple blobs or multiple objects may merge into a single blob. Models can help detect and track individual objects. (Recent Results)
Activity and Event Recognition Detection and description of events and activities from videos. Continuous event recognition. Recognize normal events and abnormal events. (Recent Results)
Lidar Processing Extract objects and descriptions from LIDAR and other range-based data.
Human Pose Analysis Human pose tracking is based on recognizing body parts frame by frame. The relationship between parts is given by a three-dimensional model and is based on the differences in position, size and appearance.
Moving Object Detection Automatic detection objects (people and vehicles) in video. Background modeling approaches allow derivation of a set of regions in the frames representing the moving objects in the scene.
Interactive Building Extraction User assisted building extraction and description from multiple images
Facial Expression Recogniton Facial Expression Recogniton system
Robust and Scalable Recognition of Objects and Events in Video (VACE) Automatic, robust and scalable recognition of objects and events in a scene using video.
GeoDec Project Description Constructing an information-rich and realistic three-dimensional visualization and simulation of a geographic location, rapidly and accurately.
ETRI: Visual Sensing for Natural Human-Robot Interaction The project aims to create and implement Stevi, a high-level vision sybsystem for use with personal service robots.
Tensor Voting This project addresses the central issue of Perceptual Grouping in Computer Vision. Over the past few years, we have developed a unified Tensor Based framework to formalize this problem. The method is not iterative, uses no hard thresholds, and there is a single free parameter.
3D Modeling of Faces from Video: Face Reconstruction From a monocular sequence, generate a dense point cloud then a 3-D triangulation. Such models improve results for recognition.
Event Recognition: Human Activities Model activities using a hierarchical activity representation.
Mosaicing and Frames Registration We present a graph-based approach merging local and global registrations into a single framework
Perceptual Interfaces for Immersive Activities (ICT) The simulation of human activity and its rendering has to take into account their interactions with the objects or people in the scene in order to create a realistic virtual experience.
Binocular and Multiple View Stereo Using Tensor Voting (IMSC) We present an integrated approach to the derivation of scene description from binocular and multiple-view stereo images, where both feature correspondence and surface reconstruction are addressed within the same framework.
Video Surveilance and Monitoring. (VSAM): 1997-1999. We present a system which takes as input a video stream, obtained from an airborne moving platform, and produces an analysis of the behavior of the moving objects in the scene.
Three Dimensional Site Modeling: Overview. -
Rapid and Affordable Generation of Terrain and Detailed Urban Feature Data MURI Project: 1996-2001 Automatic extraction of 3-D building models from multiple images and other data.
Automatic Population of Geospatial Databases (APGD) 1997-1999. Old project.
Three Dimensional Site Modeling (1999). Example results for automatic modeling system
Region Segmentation Experiments extraction. Results for Segmentation experiments on color images, finding regions as an aid in building extraction
Detailed analysis of Airport Runways
High level vision, the use of knowledge in vision.: VEIL project. 1993-1996 Veil Final Report
==================================================
Maintained by Keith Price, price@usc.edu