RE: [Nautilus-list] Help out

["Blad, John Erling" <john erling blad aftenposten no>]

First, sorry for the rotten spelling.. ;)

> Big, but probably not "core" things:
> 1. getting medusa/searching to be fast/usable/good enough for vfolders

If someone has the guts.. What about makeing an interface to Viper?
Viper is a search engine for images but I gess it can be extended to
text and most other stuff. I've heard about *grin* a system for classifying
code for reuse with similar techniques.

I haven't looked to close at Viper but I gess it comes down to turning
the client part into a bonobo-component and adding some feature extraction
algorithms to it. That is the difficult part..

To be usefull it should handle text, images, movies (sequence of images),
music.. It should probably also be able to querry remote servers on the net
if there are no local alternative.

> 5. movie file preview icons (use first frame?)

I belive the Norwegian telecom company Telenor's research fascility made
some interesting systems for tagging/viewing of movies. I think it must have
been in -90 sometimes. Maybe someone should look into that report.

> 6. those cool RSS/"smart" icons I saw a screenshot of months ago

The same telecom company also made a system to visualize voice mails.
As I recall the idea was to extract a small set of features and construct
an icon ont the fly. Features from the captured voice was used to create
parts of the icon, or a simple drawing of a head as I recall. Typical
features
was age, gender, rate of speach and length of voice mail clip. I belive age
was the size of the head, gender the hair cut, rate of speach the color and
length of the clip was the size of the mouth. 

This could really be done for other icons as well. For example it isn't very
difficult to extract tings like beat rate, volum, gender, etc from a music
clip. This could then be used to generate an icon.

John