Staying on target

facebooktwitter 

Hello, Mark My Bird lovers!

Once again, you’ve blown us away with your enthusiasm. It’s been less than two weeks since we launched, and last time I looked, you guys had collectively marked over 2,500 bills. A notable mention at this point has to go to Peregrin3, who since registering two days ago, has marked around 400 beaks. Amazing!

A few of you have asked for feedback on how well you’re doing. This is important for us too, because it helps us to identify mistakes that could affect our analyses later. Reassuringly, most of you are doing just fine. There are a few common mistakes that crop up, which we’ll take a look at below. We’re working on some updates to the landmarking tool that should flash up a warning if your landmarks start to stray too far from the brief.

How does this work then?

As we’ve already hinted to some of you who follow us on Twitter, there are some quality control measures built in to the site. One of them is that everyone who registers is given the same two training birds as the first bills they landmark: the shoebill stork (Balaeniceps rex), and the brown-chested alethe (Alethe poliocephala).

This shoebill is coming to haunt your dreams and steal your soul... © Su Neko, CC-BY-2.0

This shoebill is coming to haunt your dreams and steal your soul… © Su Neko, CC-BY-2.0

We chose the shoebill because a) it’s “easy to landmark” (big, sharp lines, sensible feathers), and b) just look at it. Look at it! It’s brilliant. We chose the alethe because, in terms of beak shape, it’s incredibly average, making it typical of most of the bills a user will see on Mark My Bird. The repeated efforts to mark these birds are our dataset for this blog. We’ll focus on the shoebill first.

I’m not going to dive in to the details of how we ran this analysis today. The method is called Geometric Morphometrics, and I’ll probably write another blog for you on that at a later date. In a nutshell, we’ll be using it to see how much variation in bill shape there is between different bird species. In this case though, everyone marked the same shape (the same shoebill… bill), so the variation will be caused by different people choosing to put their landmarks in slightly different places. For those of you who are in to morphometrics, all the analysis was done in the Geomorph R package.

Science!

Here’s what all the shoebill data you guys generated looks like:

shoepc12_rings

Each dot on this graph represents a different person, 165 in total. 85% fall inside the green circle, and have put the landmarks very close to where the members of Team Macrobird did. So well done! The person who came closest to our landmarks was hughbrazier (black dots), whose landmarks are shown compared to mine (grey dots).

shoe_hb

hughbrazier’s shoebill landmarks

What about the points that aren’t in the green circle? There are two more clusters in the graph, shown by the blue circles. These represent “loops”, where the user tried to add more points along the curves in between points they’d already placed. The point resampling along the curve gets really confused, and makes a right mess of things! One cluster is loops along the left and right curves, the other cluster is loops along the midline.

This is what happens when a loop gets put along one of the edge curves. It doesn't look much like a shoebill any more.

This is what happens when a loop gets put along one of the edge curves. It doesn’t look much like a shoebill any more, tbh.

We didn’t see this coming, but with hindsight, this is kind of our own fault. We should have been more specific in the instructions that the points along the curves have to be clicked from front to back in order. We’re working on an automatic check to try and stop people from making this mistake in the future.

I won’t name and shame the worst shoebill. Not only was there looping, but as this person looped, they also managed to mix up their left and right. Whoops!

We had slightly less data for the alethe, as 33 people had had enough, and stopped marking after they’d done the shoebill. It’s pretty much the same story, although this time, JohnF was closest to our landmarks.

JohnF's brown-chested alethe landmarks

JohnF’s brown-chested alethe landmarks

You can see that there is a small difference between where JohnF decided the keratin met the feathers and where I did. These small differences are the sort of thing we’d expect to see on a typical bill on Mark My Bird, and is why every bill has to be marked by multiple people: we need to get an idea of how much “noise” there is in the data.

In conclusion

We have some work to do to the back end of Mark My Bird, but the good news here is that most people are doing a great job. You carry on being fantastic. And if you want to see your own shoebill or alethe, post to the forum or send us a tweet with your Mark My Bird username, and we’ll try to get back to you ASAP.

Mark My Bird Top Tips

facebooktwitter 

The short version: After you’ve marked a bird, spin the beak around and look from another angle. Are the points still where you thought they were?!

This week saw the launch of our crowdsourcing website, Mark My Bird. There are nearly 10,000 species of birds for us to scan and add landmarks to, and we quickly realised that marking them all ourselves was too big a task to complete on our own. Mark My Bird was designed to ask the public to help us out with gathering the data. We’ve been blown away by the response so far, which has been incredible: over 500 bills marked in under 48 hours! So if you’re one of our landmarkers, thank you, you are a Hero of Science. And if you haven’t visited yet, head over to www.markmybird.org and get involved – you can view all of the beaks we’ve uploaded so far (including some former #BeakoftheWeek stars), and there are virtual prizes to be won too. As if marking birds wasn’t reward enough in itself…

Anyway, thanks to helpful comments and questions from our early adopters, we’ve been able to identify a few teething problems with the site that we hadn’t anticipated. In addition to the landmarking guide and extra tips sections of Mark My Bird, we’ve put together this list of Top Tips to help guide you through the most common issues (see also the Twitter tag #MMBTopTips). Ultimately, we’ll incorporate this in to the existing guides, and put together some FAQs. Don’t forget, if you have any feedback you’d like to give us then please get in touch using the Mark My Bird forum, or via our Twitter page or Facebook group. We want to hear from you! What do you like about the site? What don’t you like so much? Which birds are your favourites to landmark and which fill you with despair when they pop up? Let us know!

You guys are doing an amazing job, and what’s even cooler is being able to watch people’s landmarking improve the more bills they work on. Keep it up!

 
Top Tip #1: Spin the bill

When there are mistakes, the most common one that we’re seeing by a long way is with landmarks “falling off” the edges of a bill. Something that is surprisingly easy to forget when using Mark My Bird is that the bills you’re looking at are not flat images, but 3D surfaces. Unfortunately, this can result in edges that look fine from one angle, but when you spin the bird around to look at it from a different direction, you can see that things have gone a bit wrong.

Ooooh, lovely!

WTF?!

What’s probably happened here is that the point was accidentally placed on the surface just behind the spot the user was aiming for, and as a result, the whole curve went wonky. D’oh!

A similar problem that’s popped up with the edges is a “looping” effect, as seen here:

Loopy

What we think has happened here is that the interface told the user that they hadn’t placed enough points to finish the curve, so they went back and added some points in between the ones they had already placed. Mark My Bird is not clever enough for this! Rather than a dot-to-dot, the edge tool is more like a spider dragging a thread behind it, and if you go back on yourself, the edge follows, ultimately drawing a circle(s) between the tip and back of the beak.

 
Here, we think the user has started tracing the curve at the back and moved towards the front, rather than starting at the front and moving to the back. This effectively draws one big circle.

Going round in circles

Top Tip #2: If the midline button doesn’t work, you can place the points manually

Several people told us that the midline didn’t go exactly where they wanted it to. In some cases, you can realign the whole midline by adjusting points 1, 2 or 3 after you’ve placed them. Simply use the “prev” and “next” button to navigate to the point you want to change. You can then use the “undo” button to remove your point, or you can just click in the new location straight away if you’d prefer.

Press my buttons...

After that, click midline again to recalculate the new position.

But it isn’t always that simple. Sometimes the first three points will be perfect, but the midline go wrong and “go past” point 3, up in to the bird’s feathers, like this:

These alt puns are going too far...

We’re still not sure why this is happening and are trying to fix it! In the meantime, you can edit the midline manually, just like you can with the left and right edges, by clicking “undo”. This sequentially removes points from the back, so that you can reposition them yourself.

The undo feature is also useful on those (extremely!) rare occasions when you get an asymmetrical beak, like the wrybill, or a crossbill.

MMB user aidengill did a great job on this tricky character

MMB user aidengill did a great job on this tricky character

Top Tip #3: Bird bills are symmetrical

Using this fact can really help you to place the edge curves. Unless (see above!) you have a wrybill or a crossbill to landmark, the left and right edges should be symmetrical. Spin the bird around and have a look at it from the front, top, bottom…

Here, the user has got it correct on the left-hand side, but has accidentally followed a shadow on the right-hand lower bill, rather than the cutting edge. You can see how one side dips lower than the other when we look at the bill front-on.

In the example below, the user hasn’t followed the right-hand side as far back as the left-hand side. This could be because they’ve accidentally followed the left-hand side too far (perhaps a feather lined up with the beak and made it look longer than it is), or they haven’t followed the right-hand side far enough (perhaps a rogue feather drifted in front of the scanner). Honestly, scanning birds would be a whole lot easier if they didn’t have feathers… Either way, using the symmetry as a guide can help you to make a judgement call and correct your points accordingly.

Top Tip #4: Google is your friend

Mark My Bird has a built in Google image search button that pops up when you’re first assigned a new bird to mark. Everyone we’ve spoken to has said that having another tab open with pictures of the birds they’re marking really helps them to identify the landmark points. Plus, you get to see the rest of the bird! Yes, yes, the beaks are the best bit. But annoying as they are to scan, the feathers can be pretty spectacular as well.