A partnership between two Dunedin-based heavyweights could
have "huge" commercial potential, its backers say.
The University of Otago and NHNZ Moving Images have signed a
research agreement, with the aim to develop a cutting-edge
image recognition system, similar to those used by law
enforcement agencies.
Under the project, researchers from the university's
information science department will gain access to the
television production company's archive of more than 200,000
hours of footage.
Dr Jeremiah Deng said with the help of new technology it was
possible to identify objects in each shot, a huge boon to
those searching archival footage.
Aside from the commercial potential for those searching
archived footage, there was also consumer potential for home
users searching their computers, he said.
"The commercial potential is huge," NHNZ emerging media
manager Caroline Cook said.
At present, each shot filmed in the field comes back to NHNZ,
or other production companies, needing to be recorded, with
such things as the action, main objects and camera angles
listed on a shot list.
"This is a painstaking and laborious and very necessary job,
as the writers and directors use these 'shot lists' to create
the first rough cut of the production."
That process was labour-intensive and a costly component of
production, but an image recognition system could deliver the
right footage in minutes as opposed to days.
"Whoever developed this type of system has the potential to
sell it to other production companies," she said.
Another potential for the image recognition system, which has
been used by law enforcement agencies to detect people in
crowds, was finding the right shots from archived footage
that has not been categorised in any way.
"If we could get a system which can search for recognisable
images that haven't been shot-listed, this would help footage
archivists immensely."
Dr Deng said the NHNZ archive was a "gold mine" for the
research team, which was hoping to secure funding for the
project.
The research team will soon have a delivery of a large data
set of wildlife clips for recognition experiments, on such
popular animals as elephants, zebras, lions and polar bears.
hamish.mcneilly@odt.co.n
A name, residential address, and (preferably residential) telephone number is required from readers who comment on ODT Online. These details will not be visible to site visitors.