Finished! Looks like this project is out of data at the moment!
Thank you for your efforts! Our project is currently paused while we gather the next season of data! To browse other active projects that still need your classifications, check out zooniverse.org/projects
Also, this project recently migrated onto Zooniverse’s new architecture. For details, see here.
Results
This project has grown substantially since we launched our Zooniverse project in 2021!
Here are the major steps:
New Mexico Birds
You all collectively created over 1.5 million labels of birds at our sites along the Middle Rio Grande Valley of New Mexico. Since we had multiple volunteers sort through the imagery, this came out to about 147,000 individual birds-- one of the largest datasets of its kind in bird-world! We used your data to test a few questions:
How do experts and volunteers compare on labeling birds from drone imagery?
If you participated in this project, you know that trying to figure out what is going on in the images is no easy task, due to the "nadir" (straight-down) perspective, and at our sites, the amount of other confusing things in the environment like clumps of grass, sunlight reflecting on water, sticks, etc. We were curious to see if biologists could agree with each other on species identifications, and then whether you Zooniverse volunteers agreed with the experts.
As it turns out, when you look at individuals, there is a lot of variability in what each person believes they see in the imagery. However, when you look at the average response of each group as a whole, the experts and volunteers agreed with each other very closely-- the raw count of birds matched 91% between the two groups, specific locations of birds matched 80%, and when locations matched, identifications on duck vs goose vs crane matched 99.4% of the time-- pretty close agreement!
Here is our paper if you want to check it out!
Training the machines to identify birds for us
Using your labeled images, we have successfully trained a machine learning model to identify ducks, geese, and cranes at our original study site in the Middle Rio Grande. We currently have another paper on building that model that we will be submitting for publication in the near future and will update with more details on that soon.
Gulf Coast Birds
In 2024, we began partnering with federally-managed wildlife refuges on the Gulf Coast. We started with a visit to Laguna Atascosa National Wildlife Refuge to collect some drone imagery there and figure out what the best approach is for surveying such a large area. This winter, we collaborated with the US Fish and Wildlife Migratory Birds program on surveying whooping cranes at Aransas National Wildlife Refuge-- check back here for more data from those surveys soon!
Our goals once we have your help in labeling the new data will be to combine the new data with our New Mexico model to create a regional-scale model for identifying many types of birds in aerial imagery, than can be used by wildlife managers nation-wide.