51–60 of 88 posts
Similar: https://www.xeno-canto.org/ , a crowdsourced bird song database (over 600,000 recordings)
Edit: There are sonograms too ( https://www.xeno-canto.org/sounds/uploaded/ZWAQHOJFLZ/ffts/X... ). Could be done via image classification too.
Interesting enough to do on a boring weekend, perhaps?
Once upon a time, in the 1990's, I was in the 5th grade and the principal asked us (the class) to pull out our favorite book from our backpacks (she told us the day before to bring our favorite book) and wanted to hear the whole class reading out loud all at once.
I pulled out my "Field guide to the birds of North America" [0] and started reading along with everyone else. Most everyone else in the class either brought a comic book or some Dr. Seuss stuff. For what ever reason she noticed me and came to my desk while I was reading and stopped me and asked if what I was reading was "too hard". I don't know what I said, but I was super confused and scared thinking she'd take way my field guide and make me read dumb comics.
I have a technical question related to this post. On the front page of the site there's a prominent announcement that the site will be offline for 48 hours for a database migration. Is this fairly common, and is this reasonable? I've made some small sites but nothing that's ever gained real traction. I've always told myself that if I build something that starts to take off I should learn to do maintenance without int…
I do a bit of wildlife photography in my spare time (follow me on Insta for pictures of British birds and occasionally my cat - https://instagram.com/onion4k ). Recognising birds is surprisingly hard. A lot of them have very subtle differences that can't be spotted easily. For example, telling the difference between a common Thrush[1] and a much less common Redwing[2] is a real challenge. It's a fun and challenging p…
But actually it is true that there are a few pairs of species that cannot reliably be distinguished visually and you need to hear their calls.
Similar: https://www.xeno-canto.org/ , a crowdsourced bird song database (over 600,000 recordings)
Would it be too difficult to apply transfer learning ( or create from scratch ), an ML model to train a bird-song recogniser? Edit: There are sonograms too ( https://www.xeno-canto.org/sounds/uploaded/ZWAQHOJFLZ/ffts/X... ). Could be done via image classification too. Interesting enough to do on a boring weekend, perhaps?
Similar: https://www.xeno-canto.org/ , a crowdsourced bird song database (over 600,000 recordings)
This looks so cool. I am casually interested in bird watching, especially being able to identify them via their songs. However the moment I click to "explore" it requires a login. Yet-another login to manage before I am get to sample the data... window closed. I am not their target demo anyway.
> Yet-another login to manage before I am get to sample the data... window closed. I am not their target demo anyway. My understanding is that eBird is only for recording sightings, and not for doing identifications. For identifications you want the Merlin app/page[0]. I don't think eBird/Merlin distinguish that clearly enough. But I think what you're looking for at this point is Merlin, and not eBird. [0] https://ww…
Also, you listed the incorrect URL. The correct one is [1], which is an Android and iOS App. The page you linked to is actually the page for the species Merlin, which is a bird in the Falcon family.
Similar: https://www.xeno-canto.org/ , a crowdsourced bird song database (over 600,000 recordings)
Would it be too difficult to apply transfer learning ( or create from scratch ), an ML model to train a bird-song recogniser? Edit: There are sonograms too ( https://www.xeno-canto.org/sounds/uploaded/ZWAQHOJFLZ/ffts/X... ). Could be done via image classification too. Interesting enough to do on a boring weekend, perhaps?
A weekend, or a PhD!
The Machine Listening group at QMUL covers bird audio - see Dan Stowell's work for example: http://c4dm.eecs.qmul.ac.uk/people/dans.html
Here's a Bird Audio Detection Challenge from 2018: https://besjournals.onlinelibrary.wiley.com/doi/full/10.1111... http://dcase.community/challenge2018/task-bird-audio-detecti...
I don't know a great deal about bird song myself, but I gather the subject has lots of interesting subtleties, especially when applied to live audio or real-world recordings where it isn't necessarily known where the bird soounds appear or how many birds there are.
Sometimes I'll come across a feather when I'm out hiking and wonder what kind of bird it is from. The U.S. Fish and Wildlife service has a really nice visual database of scanned bird feathers to help identify them. https://www.fws.gov/lab/featheratlas/
That's awesome, I had no idea that existed. I feel like there's a feather-ID app waiting to happen there.
Similar: https://www.xeno-canto.org/ , a crowdsourced bird song database (over 600,000 recordings)
Would it be too difficult to apply transfer learning ( or create from scratch ), an ML model to train a bird-song recogniser? Edit: There are sonograms too ( https://www.xeno-canto.org/sounds/uploaded/ZWAQHOJFLZ/ffts/X... ). Could be done via image classification too. Interesting enough to do on a boring weekend, perhaps?
One thing that challenged my expectations was that it seems you need a lot more data than you might think. Lots of species only have a handful of samples on Xeno-Canto, and the field recordings need aggressive processing in order to isolate the relevant parts, which is hard to do that in an automated way without inadvertently destroying lots of your valuable training data.
