Description

The Words & Faces Clouds tool retrieves from the Web images associated with a given query
(a set of keywords) entered by the user, as in common Web search engines,
and then selects the retrieved images that contain faces or text (if any).

The images containing text, and the image regions containing faces, are shown as a 3D image cloud.
This tool can be used either through a Web browser, as an applet, or as a standalone application.

Try it

A demo of Words & Faces Clouds is available at this link:

Words & Faces Clouds Demo

How it works

The main window has a very simple layout:

It is made up of 5 elements:

1. a text box to enter a query;
2. a 'More' button to increase the number of images to retrieve (default: 5 images);
3. the main area of the window, where the 3D image cloud is displayed;
4. a horizontal frame, where the retrieved images and the ones under analysis are displayed;
5. a status bar where the status of the search is shown.

Keywords

After the user enters a query and presses ENTER, the search starts.
By default, this tool retrieves up to 5 images that are associated with the query,
and contain faces or text.

Each time the user presses the 'More' button, 5 new images are retrieved.

Web Engines and URLS

When the search starts, a background thread retrieves URLs associated with the input keywords
by querying several Web search engines (i.e., Google and Yahoo).

As the first URL is retrieved, a placeholder is shown on the horizontal frame to inform
the user that images are going to be downloaded.

Simultaneously a thread pool is activated to

• download the images and create their thumbnails;
• send the images to the ImageCerberus server to search for text or faces.

Images Download

When a new image is downloaded, its placeholder in the horizontal frame is replaces by its thumbnail.

A question mark is shonw over the thumbnail

to inform the user that the image is being analysed by ImageCerberus.

Server Response

If some text or face is detected by ImageCerberus inside an image

• the question mark on the corresponding thumbnail is replaced by a confirmation mark;
• the whole image (if it contains text), or the regions containing faces, are added to the 3D image cloud in the main area of the window.

Otherwise, the image is discarded, and another one is downloaded and processed as described above.
The above process is repeated until 5 images containing text or faces are retrieved (or more than 5, if the 'More' button has been pressed in the meantime).

Visual Browsing

The 3D image cloud can be used to browse and visualize the retrieved images.
The cloud is animated by moving the mouse pointer over it.
The whole image corresponding to any element of the cloud can be shown in a new window by clicking over that element.