Crowd Generated Data Labels to train your AI with Cerberus
"A quarter million game-crowd generated data labels to understand Europe's climate threat.":
One of our latest projects both to auto detect geospatial features from space, now as well to train for generative-AI, we got you covered. In various projects we have asked our crowd not to map, but they helped us to build a repository of features so AIs could take over the job while we can move the crowd to other projects.
As an example, recently for the European Space Agency we have created a large model to measure the impact of climate change on the Mediterranean of Europe. It's name was "Crowds & Machines"
In the Cerberus game client- interface we represent Earth in 3D as our own 'metaverse' whereas we use hexagons as units of analysis. With our technology we can standardize and grid entire planets!
Cerberus: Solving world problems with the power of crowdsouricng, satellites and gaming
During Crowds & Machines, the main scope was to generate data- labels to monitor durum wheat. With Cerberus we carried out over 20 sub- missions to cover the entire region in order to generate the best possible representation of all enforcing a better accuracy than what is available in the market.
The rationale is, that often for AI- training labels are generalized, but the world just isn't. In our campaign we generated labels across three important dimensions which were culture, environment and microclimate and strategically select regions to represent such diversity.
In Crowds & Machines we took the Cerberus crowd on a grand journey through the Mediterranean region visiting 20 regions. We mapped - and labelled to create a good representation of this vast region across culture, climate and geology.
Some themes Cerberus is and has been used for are:
-
Mapping Disasters so we can better plan help efforts
-
Rescuing Refugees to supply them, or getting them out of harm's way
-
Detection of illegal rainforest cut and put a stop to it
-
Supporting Farmers in developing countries to become smart and feed their communities
-
The detection of illegal fishing to protect our oceans
Some projects:
-
Cerberus: Map a Disaster (first operation, 2D mapping, award winning)
-
Cerberus: Evolution (3D mapping, immersion, safety and defense, inclusion of digital elevation- and terrain models)
-
Cerberus: Forest Falcon (introduction open data, 7 sub- projects)
-
Cerberus: Crowds & Machines (feasibility for AI data labelling)
-
Cerberus: Crowds & Machines 2 (scale up, 20 sub-projects food security EU)
-
Cerberus: Space in Africa (Impact mapping)
On Crowds & Machines, an in depth article can be read here: "A quarter million game-crowd generated data labels to understand Europe's climate threat." which was posted by our founder on LinkedIn.