Mehran Najafi

Prototype: Client-side Task Services (Java)

A task service enables a service provider to process client's data locally using generic service representatives. The developed prototype is divided into two Java Packages, as follows.

• TaskService Developer Package: This package includes APIs which can be imported and used by service developers to develop task services. The service developer designs and implements a task service by defining its task components including task model (Drools ruleflow), task knowledge (Drools rules and actions), and task data (Java beans).
• ServiceRepresentative Package: This package includes APIs which can be imported and called by the service client to generate the service representative and communication channel. The service representative has a built-in Drools rule engine.

Example: consider a task service including (i) a ruleflow as its task model; (ii) rule1, rule2, and rule3 as its task knowlegde; and (iii) an object from class Visit as its task data. Service.java represents the corresponding function to be provided as a task service. Moreover, Client.java represents a service client to invoke the service.

Face Detection using Client-side Technologies (.Net)

There are several technologies which enable client-side processing, compared in the following picture.

A face detection algorithm was implemented using different client-side and server-side approaches, as follows.

• Face Detection (C#): is a C# desktop application, installed on the client's system to process client's images locally.
• Face Detection (ASP.Net): is an ASP.Net Web application that processes images submitted by the clients.
• Face Detection (Silverlight): is a Silverlight application that sends XAML files (including skin detector + face detector classes) to a client-side Silverlight plug-in to process client's data locally.
• Face Detection (Task Services): is a web service client implemented by ASP.Net. The web service provider sends a task message to the service representative installed on the client's system. The service representative is customized based on the task components and then processes client's images locally.

Comparision of these approaches on a data set is reported here.