Protein crystallization is a difficult process where thousands of trials may need to be set up for a successful crystalline outcome. Data analytics include methods for providing conditions to be set up, automated scoring of images, protein growth analysis, focal stacking for trial images, and visualization of plates.
Sponsored by: iXpressGenes, Inc. (NIH STTR Grants GM090453 and GM116283)
Query-by-gaming is a novel way of spatio-temporal querying of video content using gaming controller. It is especially effective when objects visit the same locations repeatedly based on some actions. It is possible to provide spatio-temporal queries with a gaming controller as relevant clips based on inputs are retrieved and the database is queried concurrently.
Mosaic, sprite, or background generation from moving or tracking camera while removing the foreground objects. Applications of panorama generation include video compression, virtual camera control, effective aspect ratio conversion, and object centralization for video editing.
Sponsored by: NSF (Grant No. 0812307)
Training a classifier to recognize well-known people is possible. In many cases, as in broadcasting TV news, new faces (people) appear and it is not possible to train classifiers for people who appear rarely. We introduce how speaking people can be recognized by analyzing textual information on videos, using speaker diarization, and extracting names of people from closed captions.
Object and trajectory tracking is challenging due to necessity of proper feature points of the target. Our method can continue tracking even if the target object leaves the field of view. Moreover, our Mirage pose estimation is ana analytical method that can effectively be used for trajectory tracking with few feature points without requiring any major changes for multi-camera support.
Storing data has little use if we cannot retrieve what we look for when we need it. Multimedia data has more challenges as uncertainty and fuzziness of information to be stored, spatio-temporal context and customization of personalized queries. Our S3G semantic sequence graph, modeling of fuzzy information, determining user query based on user browsing history are some of the research studies in this area.
Multimedia synchronization involves coarse-grained synchronization among elements of a multimedia presentation. Our SynchRuler is an event-condition-action based synchronization specification where user interactions that change the course of a presentation are manipulated effectively and verified with model checking. PressBase manages synchronization through database triggers.
Multimedia networking research involves the quality of frames that are received at the clinet side especially when the network is congested. Dropping relevant packets and adapting to network conditions in time may imrpove user experience.
While enforced alignment might miss some proteins due to the necessity of similar sequences in the database, data mining methods may reveal hidden patterns in sequences. Presence of gaps and dealing with parameters such as gap penalty are some challenges of sequence alignments. Our Vector-G is an SVM-based G-protein detection method. We explain how hierarchical classifiers can be built for multi-class classification problem. We also utilize maximum contiguous subarrays for sequence alignment without requiring gap penalties.
* (NIH-through) STTR/ Fluorescence Intensity-based Scoring of Macromolecule Plates
* (NSF) REU Site: Fundamental Research Topics Related to Unmanned Systems
* (NSF) III-COR-Small: Developing Novel Mosaic Generation Methods for Object-Based Multimedia Information Systems (with REU Supplement)
* (NIH-through) STTR Phases I & II/Fluorescence Intensity-based Scoring of Macromolecule Crystallization Plates
* (UAH) Human-Robot Cooperation via Artificial Robotic Vision
* (UAH) An Infrastructure for Cybersecurity and Privacy Research in the Post-PC era