The CalPhotos image retrieval system
had its roots in a 1993-94 research project called
Chabot, which was a part of the
Database Research Group
at UC Berkeley. Chabot was developed by UC Berkeley EECS graduate student
Ginger Ogle. It used a relational database,
Postgres, to store not
only textual information about the images, but also image content.
Through the use of pre-computed color histograms, Chabot provided the ability to search for "Mostly Red" photos, or images with "White Stuff" in them.
The
images used for this research were provided by the
Film Library of the
California Department of Water Resources
(DWR), an early adopter of digital image
technology, in particular PhotoCD format. DWR continued to provide digital
photos to the Digital Library Project and by December 1998 there were over
17,000 images online. After the events of Sept 11, 2001, DWR removed the
images from online access due to security concerns.
Computer Vision Research
Chabot became a part of the Digital Library Project in 1995, when it was
renamed "Cypress" and given a web-based interface and a new database engine,
Illustra, which was developing a commercial version of University Postgres.
The image collection continued to grow, and collaboration with the
State of California expanded to DWR's parent agency the California Resources Agency, along with
its just-created information service CERES.
At the same time, researchers from the UCB Computer Vision Group (Jitendra Malik and David Forsyth and their students Chad Carson, Serge J. Belongie,
and Thomas Leung) began using the DWR
photos for their work in recognition and retrieval using image content.
They aquired a set of 40,000 Corel Stock Photos to use for this
research; these images were added to the online collection.
These images were used for testing and demonstrating Blobworld, an image retrieval
system that uses regions of color and texture to find images in
a very large collection.
Brousseau Collection of California Wildflowers
In 1996, Sheila Hurst, Outreach Projects Coordinator for CERES, and
CERES Director Gary Darling, were contacted by Brother Eric Vogel
of
St. Mary's College about a set of California wildflower slides that he
wanted to make publicly available. Using funds from DWR and the Digital
Library project, these slides were scanned and put online as the
Brousseau Collection of California Wildflowers. Later, trees and
fungi slides were added, and they now number over 12,000. The
Brousseau photographs
became the initial collection of CalPhotos.
The Biology Community
The Brousseau collection of photographs
sparked an interest in the botanical community and
led to new collaborations.
Hugh Wilson of Texas A&M's
Bioinformatics Working Group began generating links to the Brousseau pictures
from his own work with the CalFlora Database
which was being developed by Ann Dennis, then of the US Forest Service.
The Digital Library Project developed an online query and retrieval system
based on the photo database
for CalFlora, linking the photos to species information from CalFlora, and
vice versa. Interest in the
CalFlora system extended to other biological sciences and collections
as well, resulting in our development of online specimen databases
for the UC Museum
of Vertebrate Zoology (launched in 1999), AmphibiaWeb (2000), the
UC Museum of Paleontology (2002), and the Essig Museum of
Entomology (2003). The data model that was originally based on the CalPhotos database has
subsequently been used for a variety of other biodiversity research projects and is now
called the GO Model (click for more information).
The California Academy of Sciences
In 1997, we began to add the photographs of other individuals and organizations
to the Brousseau Collection. Most notably,
the California Academy of Sciences
in 1998 allowed us to include thousands of
their photographs of California plants and animals that were part of the
Manzanita Project.
The Academy continues to add several thousand new photos to CalPhotos
every year and provides funding assistance to CalPhotos.
CalPhotos
To
accommodate the new collections, we changed the name of the database from
"California Wildflowers" to
CalPhotos.
In September 2000 we introduced a new web-based image upload system that has
allowed hundreds of photographers to easily contribute new photos to CalPhotos.
Currently we receive around 1,700 new photos every month using this system.
See About Contributing Photos for more
information. Around this time we also developed an
Annotation System
for CalPhotos which allows registered experts to review
photos, add comments, and make corrections to the scientific names of photos.
Other Image Collections and Technologies
In November 1999 we made more than 1000 historical photos available from the
Museum
of Vertebrate Zoology, using TilePic
software we developed to provide zoom and pan capability for large-resolution
photos. In 2007, more than 8,000 historical MVZ photos, with TilePics, were added.
The TilePic technology is also being used for the University and Jepson Herbaria
specimen photos and for
UCMP Paleontology Collections Photos, in
particular, a collection of modern cleared leaves used for fossil identification.
In 2006, for the Moorea Biocode project,
we developed a field-based photo upload feature for specimen photos and we are extending this
to other specimen databases in the Berkeley Natural History Museums.
In May 2000 we began providing online access to 80,000 high-resolution images of
the collections of the Fine Arts Museums of San Francisco. These zoom-able
images were originally made available as part of another computer science
research project at Berkeley a few years earlier. When the Digital Library Project
ended in 2005, the Fine Arts Museum began hosting these images themselves.
As of 2008, CalPhotos contains more than 160,000 photos of worldwide plants,
animals, and habitats. CalPhotos is particularly strong in California plants
and world-wide amphibian species, but it contains thousands of photos of plants
and animals from all over the world.
The database receives more than 100,000 distinct queries each day, and serves
more than 1 million photos each day. Current usage information can be found
here: Database Query Statistics.
For more information about the photo database, see About CalPhotos.
CalPhotos
is a project of BNHM
University of California, Berkeley