Jump to content
Whatbird Community

Recommended Posts

Taken on 11-4-2021 in the late afternoon at Hansen Dam Debris Basin in Lakeview Terrace (Los Angeles) CA.

This one sandpiper was hanging out with a couple of dowitchers.  I thought it was a Least Sandpiper.  I decided to see what Merlin had to say about the photo of the single bird. 

It gave the following suggestions in order:

Semipalmated Sandpiper

Baird's Sandpiper

Least Sandpiper

White-rumped Sandpiper

 

What do you all think?

 

2021-11-04_Hansen_Dam_8129_w.jpg

2021-11-04_Hansen_Dam_8130_w.jpg

Link to comment
Share on other sites

So I have a couple of questions about Merlin.

When Merlin returns more than one possibility, are they ranked in order most likely to least likely according to Merlin's algorithms?

In the case of the above photo of the single bird, is this photo not good enough for Merlin to ID it more precisely?

Thanks.

Link to comment
Share on other sites

2 hours ago, Dan P said:

So I have a couple of questions about Merlin.

When Merlin returns more than one possibility, are they ranked in order most likely to least likely according to Merlin's algorithms?

In the case of the above photo of the single bird, is this photo not good enough for Merlin to ID it more precisely?

Thanks.

I'm =guessing= Merlin may be confused by the bird's distorted reflection.  You might try cropping the photo.  

Edited by Charlie Spencer
  • Like 1
Link to comment
Share on other sites

I thought it was a Least Sandpiper when I saw it in real life, and once I saw the photo I was sure.

I was actually trying to see what Merlin would say about some of the Dowitcher photos and was coming up with some odd results.  So I thought I'd test Merlin on the Least Sandpiper to see if it would return just a single match when its sure of what the bird is.  But apparently that backfired.

I wonder if the background color of the water is creating a problem with distinguishing the yellow legs.  And/or maybe its having trouble with determining the length of the bill.

Anyway, I always try to identify the birds first myself.  If I'm having trouble, I'll usually ask you guys on WhatBird Forum.  But sometimes if I want to get a quick opinion, I'll try Merlin.

  • Like 4
Link to comment
Share on other sites

3 hours ago, Dan P said:

Anyway, I always try to identify the birds first myself. 

I'm not the most tactful of people, and I was struggling for a polite way to encourage this.  Thanks for getting me off the hook.

I use Merlin almost exclusively to compare the bird with its database; I've never used it to ID a photo.  I don't know if I could get an image from my camera to Merlin on my phone or tablet seated at my desk, much less in the field.  By the time I accomplished that, I could have dug it out of Sibley or AAB (and I'd rather do it that way anyway) .  Fortunately, I'm never in a hurry.

For a guy that runs a network for a living, I have little experience with many consumer tools and utilities.

  • Like 3
  • Thanks 1
Link to comment
Share on other sites

On 11/11/2021 at 1:38 PM, Dan P said:

I was actually trying to see what Merlin would say about some of the Dowitcher photos and was coming up with some odd results.  So I thought I'd test Merlin on the Least Sandpiper to see if it would return just a single match when its sure of what the bird is.  But apparently that backfired.

If it couldn’t get the Least Sandpiper right, I wouldn’t trust it with Dowitchers…

  • Like 2
  • Thanks 1
Link to comment
Share on other sites

On 11/10/2021 at 12:59 PM, Dan P said:

In the case of the above photo of the single bird, is this photo not good enough for Merlin to ID it more precisely?

The photo is great. My understanding of how Merlin works (please correct me if I've misunderstood something*) is that it's been trained on (a subset of?) the Macaulay Library to allow it to look at new photos and evaluate what the species is based on what it "learned" in the training. I think the training is based on the complete pictures, as opposed to being taught to identify the legs then evaluate their colour, etc.

This is why they ask you to have the bird fill the box, to remove as much of the background as possible so it compares a greater percentage of bird to its database. Those backgrounds are part of the training though, and some of that noise is inevitably part of the id process. I think of it like a perverse mirror image of the kind of contextual clues that help us understand what a bird is: where it is, how it moves, etc.; but for Merlin that could be the angle of the sun glinting off the water or the height of the ripples from another bird bathing.

So the more confusing the bird for us, the more confusing for Merlin too -- and we have a much better chance of identifying the bird correctly because we can work through the details of an individual bird and it simply isn't designed to do so.

 

ETA: Tony L last year said something to the effect that not even the Cornell folks know how Merlin comes to an identification, which may or may not have been a joke.

 

*I did a bunch of reading last year and this is my impression from that. I'd love details on whatever I've misconstrued, if available.

 

Edited by PaulK
  • Like 1
  • Thanks 1
Link to comment
Share on other sites

On 11/11/2021 at 4:53 PM, Charlie Spencer said:

I don't know if I could get an image from my camera to Merlin on my phone or tablet seated at my desk, much less in the field.

Absolutely not your point, but the newfangled cameras can connect semi-automatically to a phone via Bluetooth and download everything. Absolute game changer. I hardly even think about the path my photos take: camera->phone->wifi network->data centre ("the cloud")->web browser->downloads folder on my computer. This is somehow easier than plugging the camera in directly.

  • Like 1
Link to comment
Share on other sites

2 hours ago, PaulK said:

Absolutely not your point, but the newfangled cameras can connect semi-automatically to a phone via Bluetooth and download everything. Absolute game changer. I hardly even think about the path my photos take: camera->phone->wifi network->data centre ("the cloud")->web browser->downloads folder on my computer. This is somehow easier than plugging the camera in directly.

I find it easier to go camera->downloads folder->cloud after I've filtered them.  I'm a very minimalist user of phone utilities and I've never had a reason to upload anything from it to cloud storage.

While my camera does has WiFi capability, I leave it and the back screen off to save battery life.  When I transfer to the computer with a cable, I can leave it hooked up to recharge the battery a bit.

Edited by Charlie Spencer
  • Like 2
Link to comment
Share on other sites

21 hours ago, PaulK said:

ETA: Tony L last year said something to the effect that not even the Cornell folks know how Merlin comes to an identification, which may or may not have been a joke.

Not a joke. Being able to understand why a specific input is categorized a specific way is an open question in Machine Learning generally (the tech underlying Merlin).

  • Like 2
  • Thanks 1
Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

 Share

×
×
  • Create New...