Recent Projects
I proposed the system framework according to the business requirements and implemented it. I divided the whole system into two parts: GUI and back-end service. For GUI part, I used Spring + Hibernate framework, JSP, JavaScrip, JDBC with Oracle database. For the back-end service, I exploited several fancy technology such as State Machine and JMX. I also used JMS to send generated journals to MQ and Scheduler to schedule back-end tasks on different states.
On large-scale software development, there are several phases and lots of people involved in. In my group in UBS, I exploited Maven, TeamCity and Subversion to manage the whole software development procedure to reduce manual work and conflicts as much as possible. Coded in Groovy.
I implemented Web Crawler, PageRank and String Matching, three most important parts of a search engine. Codes along with the detailed explanations are free to download.
There were three phases of my work. First, I improved the basic Eye Movement Model and tested it on Google Earth images with UFOs as targets. Then I used GMM to capture the patch-based distribusion and labeld the patches from new images automatically. Thirdly, I added criteria to the basic EMM to limit the search regions based on the patch "semantics".
I implemented a label tool which can help human beings easily label objects in images. And the labels are saved in a database so that a tagging algorithm can be used to cluster these labels and their corresponding objects.
I implemented an online tracking method proposed in a CVPR'05 paper by C++ plus OpenCV library. The key idea of this method is to use Adaboost to construct a confidence map for every frame, and then use MeanShift to find the mode of the saliency map so that to locate the target.