Three Million
Dear Snapshot Serengeti Community: As of yesterday, you all have made over three million classifications. That’s 3,000,000. That’s unbelievable! (For those of you new to the Zooniverse, a classification represents one person looking at one image. Or, to think of it another way, every time the “Finish” button is clicked, another classification is made.)
And, I have to admit, we really weren’t quite ready for your enthusiasm. I’m sure you’ve noticed those progress bars on the Snapshot Serengeti title page. You know, the ones that show Season 1 being done, Season 3 being almost done, and Season 2 two-thirds of the way done. Snapshot Serengeti hasn’t even been up for a week yet! The bad news is that there’s not much classifying left to do in Seasons 1, 2, and 3. The good news is there’s a Season 4 that we’re working on getting ready.
So what are these “Seasons”? They’re roughly 6-month stretches of images, based on Ali’s field seasons. Season 1 ran from June to November, 2010, and involved a lot of experimentation and damaged cameras, as Ali figured out what camera set-up would survive the animals’ curiosity.
Season 2 ran from January to June 2011. During this time, Ali gradually swapped out cameras with infrared flashes to incandescent flashes for nighttime shots. She found that the infrared night images were just too blurry too much of the time. The incandescent flashes give clearer (and color) images, with the downside that we just get one night image instead of a series of three. Also during Season 2, you can see the famed Serengeti migration in many of the images. Every year, over a million wildebeest, zebra, and Thomson’s gazelles move through the ecosystem, following the rains and new (and presumably yummy) grass. They’re in our study area from about December to April.
Season 3 ran from July 2011 to January 2012. By Season 3, Ali had mastered the camera-trap logistics. (However, we did have some trouble with a hard drive carrying lots of camera trap images that crashed. More on that some other time.)
Season 4, which will be showing up soon, covers February to July, 2012. And Season 5 goes from July to December, 2012. Yes, that’s right: the cameras are snapping away right this minute. Ali will be heading out the Serengeti in early January and will send us back the hard drive with Season 5’s images. We hope to have Season 5 ready for you all by the end of January.
So what now? Well, your amazing speed at classifying means we have the opportunity to refine our algorithm for combining classifications from multiple people. We’ve been making some assumptions about how many people need to see each image to be sure that we get the animals identified correctly. These assumptions are based on some beta testing we did, and I feel good about them. But right now, while you’re waiting for Season 4, we’re going to put some of the Season 1 and Season 3 images back in circulation for more classifications. That way, we can get an even better estimate of how many times we really need to show each image – and, in particular, how these estimates vary for easy, medium, hard, and impossible images.
So thank you for all your classifications these past six days. Please keep classifying images even when the progress bars fill up; we will be using your classifications. And we’ll have Season 4 ready for you soon.
21 responses to “Three Million”
Trackbacks / Pingbacks
- - February 23, 2013
- - February 24, 2013
Snapshot Serengeti 
Thrilled to be able to help out. I did did an ecology degree at uni, but life has steered me in a completely different direction since then. It’s great to be able to assist, and hear about the research. Its always exciting to see a ‘new’ animal (I saw my first baboons today!) and some of the photos are absolutely stunning, even without the animals. There was one particularly gorgeous one of a sunset… Any chance of us being allowed to have a copy for download? It would make a great wallpaper. 🙂
It’s on a Creative Commons License, so I believe you can copy and redistribute as long as you aren’t making money off it and give credit to Snapshot Serengeti.
You can save the picture just by right-clicking on it and choosing ‘save picture’. I have an adorable sleeping lion cub saved that way. You can also see your recent pictures in your profile.
This has been a great experience! I used the Windows Photo Gallery to check and adjust the dark photos and the fuzzy ones. So, early on I wasn’t intimidated by nearly black photos. I have also kept a slew of many of the animals to use in sketching; what better way to capture the natural movements and personalities of them as they go about their life. Models like these would be impossible otherwise! It has been so all-consuming to be doing this classification that it’s difficult to quit even into the wee hours and even when the wee hours are getting larger. I vote that we should do it all again just to be sure it’s all correct. lol Thanks for this great opportunity!
Great I have been identifying from the first day but I think it’s important that the same photos don’t get identified by the same people over and over.
They do the redundant viewing for statistical reasons. This insures that by getting a group consensus of what is actually in each image. This improves the accuracy of the whole process in a very massive way .
One of the scientists said that you see each image only once. Some images look very similar, e.g. when the grass is blowing or when a Hartbeest stands in front of the camera, breathing, for 20 minutes. I felt a lot better when I knew I was seeing each image only once.
IDing a capture of a lioness with her freshly killed zebra calf, it didn’t seem quite right to ID either animal’s activity as ‘resting’ (esp. the dead zebra), bt that seemed to be the best choice. Ideas? Then , another was of a hyena drinking–very different from eating but significant, nonetheless, don’t you think? Aside: this is really addictive, which may account for the amazing number of classifications.
I always thought that being dead and ready to be consumed means “interacting” 🙂 And I also tagged a colar on a lions neck as “human, 1, interacting”.
I had this one to classify, and didn’t know what to put. I chose ‘resting’ for the poor dead zebra as being the closest thing. Can we have a ‘dead’ tag?
I would love to be able to see the “already classified” images which have been removed from view, not to classify them but purely for my own enjoyment. I’m sure I’m not the only one who would appreciate the opportunity. Those images are probably some of the clearer ones, and some of the photos are so spectacular that I would get a kick out of looking at all of them. It is almost like being there…
Thanks for taking the time to give some background on methods, etc. Too bad there is a trade off in infra vs. incand. – better clarity bu then with a single shot, we have less info. to determine what they’re doing. All pretty much Standing with just one snapshot
Glad to have participated and surprised by how much I learned from looking at the animals. African nature shows on television will never be the same. Eager for Season 4 & 5 to start. Could use an “Ear” photo identification tab for all those close up head shots.
thanks for the opportunity to help classify . I learned a lot ! its quite addicting and when the screen said “Were Done” it was a bit of a letdown. Can’t wait for more images to classify and yes we need a Dead button also more choices in Birds would be good . Thanks again
Thanks for the feedback. I’m amazed too at the speed with which this work is being done. Well done everyone! As others have said, it is very addictive. I’m a former Serengeti scientist myself, so most of the time I know what I’m seeing, but even so there are many challenges and surprises. The remote cameras capture views (and even species) which you would never see if you were out there driving around. Some of the images are AMAZING and I really hope there will be a book or online gallery featuring some of the best.
The ‘looks like’ images are not the best. I think one facing, one side on and one from the rear would best help us chose the right animal.
Some of the ‘capture’ images show marking, horns etc., far more clearly than your labrary snaps.
Could there be a box like the ‘young included’ that we could tick to flag an outstanding image that you could transfer to that animals labrary picture set?
I have enjoyed working on this so much that I was genuinely sad to see that we had reached the end, if only for now. I have absolutely loved seeing all the photos and as mentioned before, some are nothing short of stunning. It is hard to believe that some of them were not specifically staged for their amazing composition and content. I cannot wait to see the next season and take on all the challenges that will entail. Thank you so much for the opportunity to participate.
Reblogged this on Alaskan Librarian and commented:
I’ve played around with a number of these “citizen science” projects and I think this is the fastest an initial project phase has been completed. I think it means that people really enjoying looking for animals.
It would be great to add some more choices of birds to classify. You could just rerun all the images that have already been classified as “Bird (Other)”.
I think that would also be very enjoyable for the classifiers, because they’d be guaranteed to get some images with an interesting bird, as opposed to some waving grass or yet another zebra/wildebeest/Tommy mixture.
Just seen your comment regarding a crashed hard drive. I’m sure you have lots of technical help getting images off such a thing, but I do recommend this program: http://www.cgsecurity.org/wiki/PhotoRec …I used the companion program ‘TestDisk’ to retrieve an entire PC drive full of video/ graphics work. http://www.cgsecurity.org/wiki/TestDisk
The down side is that each file is given a serial number rather than name, but this works when things are really badly corrupted. Hope this helps.