Archive for the ‘Rules and Scoring’ Category

“Dashboard” Feature

Sunday, February 15th, 2009

I’ve completed enough of the “dashboard” page I spoke about earlier to go ahead and post the URL to it, if you want to check it out. It’s at:

Currently it has the following:

  • Thumbnails of the last five photos taken today.
  • Thumbnails of the top five photos taken today (judged by the number of “favorites” stars assigned to it by users other than the person who took the photo).
  • A table ranking the top players today.

The player rankings are being done by a different system than the official CONE Welder scoring (though it probably has some elements in common). For the purposes of the dashboard page’s player rankings, I do the following: For each player, I compute a raw score, and a ranking against the other players based on that score, in each of three categories: correct IDs assigned today (with “correct” being in terms of CONE Welder’s idea of a correct ID, which I believe is that at least 2/3 of the people entering an ID have to agree), species photographed today, and favorite points (the stars) assigned to that user’s photos taken today. For each of those three categories, a higher raw score is better, for ranking purposes.

I also compute a combined score, which is the sum of the player’s rankings in the other three categories. I then rank the players based on their combined scores, with that ranking being what determines the overall order in the table. Note that for the combined category, you want to have the lowest raw score in order to rank highly.

The data is refreshed every 15 minutes during the daylight hours, and every hour at night. The actual times of the daytime updates are 5, 20, 35, and 50 minutes past the hour.

I have a bunch of ideas for things I’d like to add, including:

  • The ability to view results for any day, not just the current day.
  • The ability to view a similar page where the computation is based on all-time data, not just a single day’s data.
  • Pages for each user in the game, with that user’s most-recent photos, top-rated photos, and a list of species photographed.
  • Pages for each species in the game, with (again) that species’ most-recent and top-rated photos, as well as a graph showing number of identifications of that species per day.

In general, I want everything to be clickable, so it’s easy to navigate your way around and drill down to look at interesting data.

Please take a look, and let me know what you think. In particular, if you have problems or questions, or if you have suggestions about which features you’d most like to see added, please let me know. Thanks!

Idea for a New Scoring System

Sunday, February 8th, 2009

One of the people on the team at CONE, Zhang Yan, has been kind enough to work with me recently on setting up an interface that allows realtime querying of game data. My plan is to use this to build a “dashboard” page of sorts that ranks players according to a different system than the one the game currently uses.

I hoped to get the initial version of that up and running today, but unfortunately I’ve had other commitments that have kept me from working on it. As a result, I think it’s probably going to be another week or so before I’ll have the initial version of that ready.

In the meantime, though, I wanted to give an idea of how it would work.

Basically, I’m thinking of displaying a table that ranks the players that have been active in a given day according to three different criteria:

  • number of species that player has correctly identified in photos taken that day
  • number of favorites points (the stars, I mean) that have been assigned by other players to that player’s photos taken that day
  • number of identified species in photos taken by that player that day

In each of these categories, each player would be ranked relative to all the other players who had received a ranking that day, and would get an ordinal score (first, second, third, etc.) based on that. The three different ordinal rankings would be combined to give an overall ranking for each player.

My hope is that this will allow players to pursue different strategies in a way that will be fun. A player could use his or her 10 shots to try to photograph as many species as possible. Or to take the most interesting or artistic images possible, in an effort to win the favorites points ranking.

Yan has set up the interface to return the necessary data. I just need to build the pieces on my side that query that data and display the results on a web page.

“Bonus” Zoom (aka “Wide Angle”)

Saturday, November 1st, 2008

Today idbirds and I were chatting about something we’d noticed from time to time: How it apparently is possible to get a view with the camera that is zoomed out farther than is normally allowed. We did some experimenting, and it turns out that it’s pretty easy to trick the system into giving you such a zoomed-out view. Here’s a photo I took when the camera was in what we’ve taken to calling “bonus zoom” mode:

Image 106894

(Update: abirch suggested calling it “wide angle” mode, which I think is probably a better name.)

It seems like this view would be handy in certain circumstances. It helps avoid some of the “tunnel vision” I was talking about in a recent post. You need to be careful with this, since it will tend to make small birds disappear in more-distant views. I’m sure I’m not the only person who’s noticed how smaller birds can seem to vanish when the camera is zoomed out too far, especially when those birds are located far away. That’s why I tend to zoom in a bit when scanning distant locations: There’s no point in looking for birds if you’re zoomed out too far to see them.

To give you a sense of scale, there’s a Golden-fronted Woodpecker visible on the fountain in the lower lefthand corner of that bonus-zoom shot above. It’s pretty tiny, even though it is quite close to the camera. So don’t expect to see small birds in bonus-zoom mode if you’re pointed very far away.

But I think bonus zoom could be useful when looking at a field of view that is relatively close to the camera, waiting for something to fly in that can then be zoomed in on for a closer view. The bonus zoom lets you keep more area under surveillance without having to constantly be panning back and forth. Also, if you could pan while in bonus zoom mode, you could cover the same field of view more quickly, making fewer stops.

There are two parts to this bonus zoom trick that we worked out today:

Trick #1: Getting into bonus zoom mode

This takes two users working together.

The system normally restricts how large a field of view you can request. When drawing a rectangle on the panorama, or when using the minus sign (-) button to expand your most-recent request box, you can’t go past a certain-sized box. (I’ll refer to this normal maximum-zoomed-out view as “max zoom”.)

It turns out, though, that the “game” interface we use to issue requests artificially limits this max zoom. That is, the underlying camera system is capable of zooming out farther than that. In particular, if two users each draw a max-zoom box slightly off-center from each other at roughly the same time, the system will try to accommodate their requests by zooming the camera out a bit more to create a view that encompasses both requests. As long as no one then pulls the camera back by issuing another request, the camera will stay in bonus-zoom mode.

The specific way we accomplished this today was this: One user would start drawing max-zoom requests on the panorama, centering those requests on the red hummingbird feeder to the left of and slightly above the fountain. (There’s nothing magic about that location, but it made for a convenient landmark as we coordinated in chat.) At the same time, the second user started drawing max-zoom requests that were centered about halfway between the hummingbird feeder and the upper-righthand corner of the first user’s request box.

In other words, the second user’s request box was positioned in such a way that its upper righthand corner was above and to the right of the first request box by an amount about half the size of a normal max-zoom box. In still other words, the finished bonus-sized box ended up being about 50% larger than a normal max-zoom box. From the small amount of experimenting we did today, it looks like that’s about as big a box as we can make. When we tried to make a larger bonus-zoom box, the trick didn’t work. That probably reflects a hard limit in the camera’s underlying control system.

The trickiest part of this is knowing when to stop drawing boxes. If you watch the panorama carefully, you can see when the camera creates a bold box representing bonus-zoom mode. Once you see that, it’s important that you stop issuing requests. If you (or another user) draws another request after that, the view will shrink back to that size, dropping you out of bonus-zoom mode.

It seems to work best having just two users do this trick. When we tried it with a third user, we ran into problems, since the chances increased that a late request would drop us back out of bonus zoom after we’d achieved it.

Trick #2: Maintaining and restoring bonus zoom

As you can see, it’s a little bit of work getting into bonus zoom. Also, once you’re in it, you can’t issue another request (zooming or panning) without dropping the camera back to a normal view. But we figured out the following trick: If, once the camera is in bonus-zoom mode, you leave the game and immediately return, as you arrive you will be given a default black request rectangle that is equal to the camera’s current view. In other words, you’ll get a request rectangle that is bonus-zoom sized.

This is really useful. If you draw a rectangle on the panorama, or use the zoom buttons (+/-), you’ll get a normal-sized request box, and lose your bonus-zoom request box. But as long as you limit yourself to using only the arrow keys on the camera controls (up, down, left, right), you can drive the bonus-zoom request box wherever you want. In other words, you can pan around the panorama in bonus-zoom mode.

Besides being able to pan in bonus-zoom mode, there’s another advantage to getting a bonus-sized request box. If a bird appears, and some other user zooms in on it, you will keep your bonus-sized request box. As long as you don’t issue any other pointing requests, you can easily restore the bonus-zoom mode afterward by nudging your request box with the arrow keys.

Did you follow that? Once you get into bonus-zoom mode using trick #1, one or more users can then obtain a bonus-zoom sized request box using trick #2. They can then pan around in bonus-zoom mode using the arrow keys, and, if a bird is seen and the camera is zoomed in by another user, they can return the camera to bonus-zoom mode afterwards by nudging the arrow keys.

CONE Welder Videos at YouTube

Sunday, October 19th, 2008

I was in the game for a little bit this morning (something I haven’t been doing enough of lately; shame on me), and had the fun of meeting a new user, amy0321. She (I’m assuming she’s a she, based on the username) mentioned that she’s a graduate student doing research on the technology of wild bird observation, and that she’s located in Beijing. She also mentioned that she’d seen some videos on YouTube about Professor Goldberg and CONE Welder, which led me to check them out. They’re pretty cool!

Here’s one I’d seen before, though not on YouTube: It’s a demo of how to use the site, with narration by (I believe) Bryce Lee:

Here’s a short (53 second) video of Prof. Goldberg talking about CONE Welder to (I believe) a classroom of younger children:

Here’s a longer (1h 26m) video of Prof. Goldberg talking about various projects he’s worked on, including CONE, in a presentation at Stanford University:

Finally, here’s Prof. Goldberg in an hourlong interview from the series “Conversations with History”, talking about “his dual careers as an industrial engineer who designs robots and an artist whose creations use robots to stimulate understanding of technology’s impact.”

Like I said, really cool stuff! Thanks, amy0321!

Cone Welder Scoring

Friday, October 3rd, 2008

This morning several CONE players chatted at length about the way our scores went up and down, for no apparent reason. The discussion showed us that many people were experiencing the same yo-yo scores, and that it was incomprehensible to all of us. We agreed that the information about scoring available to us in the Tutorial did not adequately address our questions. Related to that, some of us have read all or part of Bryce Lee’s paper, and are increasingly aware that we do not know the whole story with this “game.” We agreed that I would write to Ken Goldberg and ask for his comments. Our correspondence follows.

10/03/08 8:28am Loughman to Goldberg: There was a lot of discussion on the CONE-Welder Chat this morning (7:31-8:21 am) about scoring for the game. Would greatly appreciate your reading it, sharing with appropriate colleagues, and getting back to me with your comments which we could post on the Blog. Thanks.

10/03/08 9:18am Goldberg to Loughman, with copies to CONE Welder team: Please rest assured that we are definitely not playing mind games with CONE-Welder players:

Over the summer, Yan observed his score going down a few times and we worked really hard to track down a potential bug in the system but found it hard to repeat. Scores can go up when one is offline due to others classifying your photos, but they should never decrease unless in the relatively rare situation where a photo gets re-classified.

We apologize for this bug. We’ll take another look at this and try to fix it. We have utmost respect for our players and always strive to maintain fairness.

10/03/08 9:51am Loughman to Goldberg, with copies to CONE Welder team: Thanks for your quick response. Fixing the bug would certainly alleviate SOME of the angst!

But the other thing we need is for someone to post clear and complete *rules of the game*. Bryce Lee’s paper does seem to indicate that Big Brother is involved. How? As I said in the Chat, that’s creepy!

If we’re better informed, we can make more meaningful contributions.

Your further comments appreciated,

Two That Got Away

Saturday, December 1st, 2007

I’ve mentioned before how impressed I’ve been by the way identification consensus emerged around some of the trickier shots in the game; by and large I can’t think of any significant miscarriages of justice.

The only real regret I have about the identifications made during the game were these two species, neither of which ever got enough votes to qualify for the ID that I think the images deserved. (The Yellow-rumped Warbler would have been on the list, too, except that one of the handful of shots we got of that bird made it to an ID.)

First up are these two shots, taken at 11:06 a.m. on October 26 by petemokazfi and robin54, respectively:

Image 40327

Image 40328

I’ve written about them before, and the more I look at them the more convinced I get that they really do show an Orange-crowned Warbler. Sadly, we never got the 2/3 majority we needed to get the ID. For the first image, the vote count we ended up with was:

7 Orange-crowned Warbler
3 Lesser Goldfinch
1 Warbling Vireo
1 Wilson's Warbler

For the second image, the final tally was:

8 Orange-crowned Warbler
3 Lesser Goldfinch
1 Warbling Vireo
1 Hooded Oriole

So, we would have needed 3 more votes to flip the first one, and 2 more votes to get the second one. But with voting closed with the end of the game, that will have to remain an unrealized promise.

Likewise with this image, which was taken way back on May 7 at 1:58 p.m. by bluebean:

Image 7128

I wrote about this image before on the blog (see Well, Hello Little Lady), and again, the more I look at the image the more likely I think it is that that was a female Western Tanager. The final vote tally for this image was:

14 Western Tanager
4 Lesser Goldfinch
3 Bullock's Oriole
1 American Goldfinch
1 Summer Tanager
1 House Wren
1 Pygmy Nuthatch

So we would have needed 8 more votes to flip that one.

Oh well. Better luck next time.

noho_bird_club’s Big Little Big Day

Friday, November 30th, 2007

Several times I tried to get 10 distinct identified species in my 10 shots in a day; it was always a lot of fun, even when I didn’t quite make it. Now that I can play with the metadata, here’s the record of the top “Little Big Days” on the system, ranked (first) by the number of species identified, and (second) by the time of the last shot of the day. That is, if two users both got the same number of species, the “winner” was the user who got their last shot earlier in the day.

Anyway, here’s the top 20 performances on the system judged by those criteria:

| user           | begin               | end                 | count |
| noho_bird_club | 2007-05-18 05:51:10 | 2007-05-18 12:06:25 |    10 | 
| elanus         | 2007-05-12 05:53:19 | 2007-05-12 16:06:32 |    10 | 
| elanus         | 2007-05-05 06:03:41 | 2007-05-05 16:55:51 |    10 | 
| kitcat         | 2007-08-27 08:09:52 | 2007-08-27 17:00:01 |    10 | 
| vireo          | 2007-10-19 09:12:31 | 2007-10-19 17:43:39 |    10 | 
| elanus         | 2007-08-12 09:11:08 | 2007-08-12 19:32:43 |    10 | 
| noho_bird_club | 2007-05-25 06:23:56 | 2007-05-25 12:18:10 |     9 | 
| robin54        | 2007-10-26 07:18:31 | 2007-10-26 13:15:46 |     9 | 
| elanus         | 2007-10-19 07:38:19 | 2007-10-19 16:08:34 |     9 | 
| vireo          | 2007-10-26 16:25:52 | 2007-10-26 17:14:26 |     9 | 
| kitcat         | 2007-10-21 11:54:49 | 2007-10-21 17:52:17 |     9 | 
| vireo          | 2007-10-17 13:26:50 | 2007-10-17 18:16:40 |     9 | 
| noho_bird_club | 2007-05-15 06:38:32 | 2007-05-15 19:27:54 |     9 | 
| elanus         | 2007-10-17 07:35:23 | 2007-10-17 08:46:13 |     8 | 
| noho_bird_club | 2007-06-01 05:51:48 | 2007-06-01 11:20:14 |     8 | 
| spurdin        | 2007-06-12 08:59:27 | 2007-06-12 15:11:12 |     8 | 
| elanus         | 2007-11-07 13:35:32 | 2007-11-07 16:07:46 |     8 | 
| robin54        | 2007-10-20 07:58:33 | 2007-10-20 16:35:22 |     8 | 
| birdbrain      | 2007-10-11 09:41:26 | 2007-10-11 18:03:16 |     8 | 
| birdbrain      | 2007-08-10 13:44:09 | 2007-08-10 18:45:09 |     8 | 
20 rows in set (0.22 sec)

As you can see, there were 6 times that a user went 10 for 10; I’m proud that 3 of those were mine. (It probably helped that I was the only person trying to do this 10-for-10 thing most of the time.)

But noho_bird_club is the champion, hands down, in terms of earliest completion of a perfect-10 day: On May 18 he got his 10th shot of a distinct species at 12:06:25 p.m., more than 4 hours earlier in the day than my next-place finish.

Here are the 10 shots he got that day, in order:

American Robin at 5:51:

Image 10419

House Finch at 6:29:

Image 10420

House Sparrow at 7:35:

Image 10437

Pygmy Nuthatch at 8:36:

Image 10444

Chestnut-backed Chickadee (small, but clearly identifiable) at 8:43:

Image 10449

Western Scrub-Jay at 8:48:

Image 10457

Anna’s Hummingbird (a little fuzzy, but again, definitely identifiable) at 8:52:

Image 10461

Mourning Dove, shyly preening but no doubt on the ID, at 10:56:

Image 10517

Several Rock Pigeons on the distant roofline; probably the sketchiest photo of the bunch, but again, for birdcam regulars there’s no question that’s what they are, at 11:42:

Image 10526

And finally, this cute shot of the Dark-eyed Junco looking at us in its reflection in the thistle-seed feeder:

Image 10530

Congratulations, noho_bird_club!

noho_bird_club Back on Top

Tuesday, June 12th, 2007

The seesaw battle for the top of the leaderboard continues, with noho_bird_club taking the top spot back from fingerlakes. Here are the standings as of a few minutes ago:

Congratulations, noho_bird_club!

Congratulations, fingerlakes!

Saturday, June 9th, 2007

I spoke over-broadly about noho_bird_club kicking all our butts in scoring. He’s kicking all our butts except one: fingerlakes’. Because fingerlakes, as of this evening, is the new top-scoring user in the system:

Congratulations, fingerlakes!

Classification Speed Up!

Monday, June 4th, 2007

I mentioned in an update to a previous item that I no longer think the system has been intentionally nerfed. From some email correspondence I’ve had with the CONE SF system’s creators, I now believe that all the slowdowns I’ve been perceiving as intentional actions are in fact just garden-variety slowness resulting from performance issues with the site’s database.

In an email I got early this morning from Bryce Lee, who is listed on the site’s credits page as being responsible for “Database and Website Design and Engineering”, he mentioned that “tomorrow, we are rolling out the classification fix and will be addressing the my gallery shortly.” So I popped into the system just now (at 6:30 a.m. Pacific time) and entered a couple of classifications, and the result page popped right up within a second or two. That’s awesome!

Go, Bryce! 🙂