[Tweeters] Options for photo IDs - Canyon Towhee ID on my ID request

Hal Michael ucd880 at comcast.net
Wed Dec 9 07:16:04 PST 2015

We seem to believe, or at least are told, that technology will provide all the answers. A common comment that was in (at least older) field guides is that some birds are not identifiable in the field. While this was generally applied to Empids and maybe immature gulls, it has broader applications.

One particularly humbling experience for me occurred with fish. At a conference, one guy showed a series of fish pictures and asked us to identify them as coastal cutthroat or rainbow trout. We were in close to 100% agreement as to which were cuts and which were 'bows. Then he told us were were close to 100% wrong, based on the genetic identification run on each of them. A few years later this happened to a colleague working with some fish in Lake Washington.

Even the computer programs like Merlin are based on averages and will miss outliers, odd plumages, and different poses. They will also miss size. In this case, unless it was a bantam towhee, the size differential between a towhee and a titmouse should separate them immediately. They are just very different in size and shape.

The last aspect, commonly used by seabird observers, is what is called "jizz". As explained to me, jazz is simply the sum of behavior, flight pattern, shape, and a myriad of other aspects that are accumulated by actual observation. Species A has deep wingbeats, species B has a "twinkling (see swifts)" flight, species C is skimming the surface, and Species D is 5m above the waves. Again, in this case, the towhee is on or near the ground and the titmouse is up in a tree, probably near the ends of branches.

This is what makes observing nature, and not just ticking, so much much fun. The more I watch things, the more I learn about them and the easier they become to identify.

And I still get IDs wrong and sometimes it will amaze me how far odd I can get.

Hal Michael
Science Outreach Director, Sustainable Fisheries Foundation
Olympia WA
360-791-7702 (C)
ucd880 at comcast.net

----- Original Message -----

That is interesting, I will have to check that out. Much like facial recognition for humans I guess, seems like a logical thing to do. I will try this with some other birds this evening, the more we try to use this the better it will get.

Jane Hadley
Tuesday, December 8, 2015 11:21 PM
Bill - Somewhat confusingly, Cornell has two different Merlin offerings. One is the original Merlin Bird ID app that is available both online and for mobile, as you mention. With this program, you answer questions about where and when you saw the bird, the color, shape and size of the bird, habitat, etc. and it guesses what your bird is based on your answers.

Their other Merlin program is called the Merlin Photo ID. It is at the link I gave in my earlier post: http://merlin.allaboutbirds.org/photo-id-help/

At this link, you upload your photo, draw a tight box around the bird, click first on the tip of the bill, click second on the eye, and click third on the tip of the tail. Then click on the submit button, and the Merlin program will guess what your bird is. It often is right, but also a fair number of times wrong. The better your photo, the better Merlin performs, which is not surprising. It also apparently does better when the bird is in a "typical pose." The Cornell researchers are trying to improve Merlin's performance and I gather that the more photos it sees, the better it becomes.

One of the shortcomings of the program is that it can ID only the commonest North American species -- 400 in its database. So if your bird falls outside of that group, you're out of luck. But this is not a criticism -- it is a miracle that Cornell has gotten as far as it has with this program.

As I mentioned earlier, just for fun, I ran one of your photos, Bill, through Merlin Photo ID, and it identified it as Canyon Towhee.


On 12/08/2015 07:47 PM, retief wrote:

Tuesday, December 8, 2015 7:47 PM
First of all, thank you all so VERY much for helping to educate me, hopefully this old dog will retain a bit of your knowledge and help.

These are great suggestions for searching for ID. One thing I find difficult is that even when someone gives me an ID, as in this case, I look in my Sibley's and what I see does not quite look right, although close. The reply from Hal Michael really helps me to understand this more completely, and his "single point in time" comment resonates with me. I try to got through these steps prior to asking, but sometimes they don't quite get me there. I have downloaded the Merlin app on my phone and tried using that, but Canyon Towhee did not show up in the list, not did it via the Online version. Jane, can you tell me what options you used so I can figure out how better to use the app?

This is why I find Tweeters to be such a great resource, especially for people like myself who are do not fit what I think is the typical "birder", as my "focus" (bad pun intended) is a bit different. That being said I do want to be as accurate as possible, and you folks simply all deserve a very big "Thank You" from you, so ..

Thank You very much.

Tweeters mailing list
Tweeters at u.washington.edu
Jane Hadley
Monday, December 7, 2015 1:17 PM
Bill Dewey asked for help identifying a bird he photographed in New Mexico. I also believe that it is a Canyon Towhee.

Several people raised questions about whether Bill's pictures matched pictures in their field guides.

I wanted to suggest that one helpful option for matching bird photos is Google Images. If you go to https://www.google.com/imghp and type in "Canyon Towhee," you'll be shown scads of pictures of Canyon Towhees (plus sometimes other birds thrown in thanks to the vagaries of the algorithmic approach). You can even search by sex, age and season: "female Hooded Merganser" or "Juvenile American Robin" or "basic plumage Golden-crowned Sparrow," for example.

When I did the "Canyon Towhee" search at Google Images, I saw that the Canyon Towhee sometimes shows a crest (5th picture on the first row).

Another sometimes helpful option is Cornell's Merlin Photo ID. This is at: http://merlin.allaboutbirds.org/photo-id-help/

I ran one of Bill's photos through this program and it did identify it as a Canyon Towhee. However, this program is in "Beta" right now and is somewhat hit-and-miss in its accuracy. Still, I've found it to be fun to try and easy to use.

Jane Hadley
Seattle, WA

Tweeters mailing list
Tweeters at u.washington.edu


Bill Dewey

Tweeters mailing list
Tweeters at u.washington.edu

More information about the Tweeters mailing list