1,000,000 Classifications!
-
by DZM admin
I didn't want to let this go by without mentioning it, but we just passed a whopping 1 million classifications on Chicago Wildlife Watch!
Incredible work, everyone. This project has become a booming success, and it looks like we're still only about 40% of the way done.
Thank you for everything... keep up the fantastic work!
I think that the science team also has some bonus stuff planned for everyone to celebrate 1 million, so stay tuned!
Posted
-
by mason_UWI scientist
Hooray! Yes, we have a really great post coming up next week for everyone. Thanks for all of your help!
Posted
-
by llehrerlpzoo.org scientist
Amazing! Thanks to all for your hard work!
Posted
-
by mason_UWI scientist
Posted
-
by DZM admin in response to mfidino@lpzoo.org's comment.
Absolutely fascinating. Thank you so much for sharing.
So cool to see all of our classifications translating into real data maps! 😄
It would be especially cool if this could be incorporated into a zoo exhibit somehow... ?
Also, what about deer? Is there a reason for no deer map?
Thanks so much!!
Posted
-
by mason_UWI scientist
Ah, you caught me! I excluded the deer due to a small coding error. Currently, I only have 118 site coordinates on my computer (we have more sites than that and I need to update our database). The 'high' activity deer sites are those that I do have the coordinates for so it plotted out a little bit differently. Here is what the deer one looks like.
Posted
-
by escholzia
Interesting post! Will you share what you decide to do about the misclassified pics? Working with messy databases is challenging - between the newbies and the careless oldbies (yo!) and the #whatsits, you'll have all sorts of data quality issues to deal with. Cleaning that up without your staff having to eyeball thousands of pics will be a challenge. If you focus on fixing outliers (a skunk on the waterfront?), is that cherry-picking?
Posted
-
by Bonnie123 in response to mfidino@lpzoo.org's comment.
Thank you so much for this update! It is always nice to see that what we are doing is useful and has a purpose. Thanks again!
Posted
-
by mason_UWI scientist
Dealing with user bias is actually a very interesting question, and I will very likely approach it in three different ways.
- Majority vote (the current process)
- Determine individual user ability, and give a photo the maximum probability from the best user. (i.e. if Tom has a 99% success rate for coyotes and says a photo is a coyote, it matters very little what Dick and Harry enter).
- Bayesian modelling! Probably the one I am the most excited about, but will be the most intensive process.
The best technique will be a trade off between speed and accuracy, so if the second process provides results similar to the third process it makes sense to just do the second. What will be interesting is to verify all of this, as currently we do not know what the best process will be. Furthermore, when we start using these data in statistical analysis we will need to include extra parameters to account for what we call 'false positives' in our data set (i.e. saying a species is at a site but it actually is not). With most ecological data that scientists collect we tend to assume that we do not make false positives, which is an issue on its own, but with the inclusion of citizen science data it will be important to account for this error.
I should hopefully have some time to start playing around with these data over the summer, and I will keep everyone updated on the process!
-Mason
Posted
-
by ElisabethB in response to mfidino@lpzoo.org's comment.
Thanks for the background info ! It is always fascinating !
Have you thought about getting in touch with the Serengeti scientists ? I can imagine that they have similar issues dealing with our classifications.
Posted
-
by mason_UWI scientist
That they do, and yes we have, plus some other folks around Chicago.
Posted