WAAROM DETERMINATIES OP BURGERWETENSCHAPPLATFORM INATURALIST ERTOE DOEN

Our Diopsis cameras use the Intel Geti platform for computer vision AI, and were featured as a use case by Intel, with a quote from our Chantal Huijbers. From crop pollination to serving as a food source for other organisms, insects
play a vital role in our ecosystems.
https://ddxcf69zoanj.cloudfront.net/documents/geti-naturalis.pdf
https://geti.intel.com/case-study/naturalis-and-diopsis-harness-the-power-of-the-intel-geti-platform-to-overcome-data-challenges-in-biodiversity-research
However, several studies have shown insect
populations are declining at a concerning rate. A 2019 study found dramatic rates
of decline may lead to the extinction of 40 percent of the world’s insect species
over the next few decades.1
A study from 2017 found over a 75% decline in the
number and diversity of insects in protected areas of Germany in just 27 years.
2
To discover and address which pressures are most responsible for this decline,
ecologists need an efficient way to continuously monitor insect populations over
a long period of time. Traditional methods of manually counting and identifying
insects are prohibitively expensive and labor intensive, which is how the new
Intel® Geti™ platform combined with digital technologies comes into play.
The DIOPSIS Consortium was founded in 2018 to explore ways to automate
monitoring. This collaboration between Naturalis Biodiversity Center, EIS
Kenniscentrum Insecten, COSMONiO, Radboud University, and Faunabit led to
the creation of DIOPSIS, a fully automated system that uses smart cameras and
AI to photograph, identify, and track insects.
Specialized insect cameras were developed for this project, and 100 of them
were deployed across natural, urban, and agricultural areas throughout the
Netherlands. Images were captured every 10 seconds, 24 hours a day for 4 to 8
weeks per year, resulting in a dataset of more than 15 million images and tens of
millions of individual insects.
Processing this extensive data set the traditional way was out of the question.
Typically, an expert might detect and identify 500 specimens per day, at most.
One season of data would take 20,000 days of expert processing, not including
the time needed for measurements and biomass estimations. Finding a better
processing method was crucial.
The DIOPSIS Solution: Automated Monitor
https://ddxcf69zoanj.cloudfront.net/documents/geti-naturalis.pdf