BirdWeather Site

[timsterc]

Hi Everyone!

I emailed Patrick yesterday - thanking him for his work on BirdNET-Pi, and introducing myself and my project.

... I had gotten turned onto the BirdNET app earlier this summer, and then stumbled on Stefan's BirdNET-Lite repo on Github. I started off with a local recorder in my backyard (Moss Beach, CA), processing files with a script, but wanted to get something real-time, and had excellent timing as Patrick had just released BirdNET Pi!!!

I've since gone a bit crazy ... I had a vision of being able to see multiple stations, with statistics across a region. We've created a web app - https://app.birdweather.com/ - it's live, with my backyard Pi station posting detections/soundscapes, along with a load of audio streaming (Locusonus) sites, and YouTube Live Cams that are getting processed by my Mac Mini M1. I've also done a few drop-off(s) where I've left a recorder for a few days. I then run a script to POST the results to the BirdWeather site. The stations are color coded based on the type. The Top Species are responsive to zoom, and the selected time region.

Still a work in progress, but I'm finally happy enough to start sharing :-)

If any of you have any interest in putting your station on BirdWeather, let me know, and I'll create a station and assign a Station ID for you. I sent Patrick my (modified) analyze.py script which includes the API hooks for the BirdWeather server ... I'll wait to see how it's best to incorporate this into the BirdNET-Pi world.

I'm also working on an iOS/Android app as well, with a view of people being able to use that and a wireless BT mic as a way to get a 'station' online for those that aren't so savvy.

Full disclosure ... I'd originally toyed with the idea of building something for commercial release, but having learned about Haikubox (and the recent change in license) figured that my efforts would be best put towards a non-commercial data visualization app. I have to say, it's been the most fun I've had coding in a while!

Happy Holidays!

Cheers!
-Tim

[mcguirepr89]

Hello again! So glad you have posted this here -- I am going to echo this post on the Testing discussion to be sure those folks see as well.

Please go ahead and create the station information for my installation and I will begin some testing :)

Edit/Update:
I have been thinking about this a bit and, without knowing your Station Numbering schema, I wanted to suggest that stations take on a schema similar to a standard IP telephone directory. That is, they should all be 4 digits (or 5, for a more ambitious global station goal), with the first digit indicating the Continent, leaving the last three digits for standard station indexing. This will allow stations to be grouped by continent, while also still indexing as expected.
For instance,

1001 could be the first station in North America
2001 the first in Europe
3001 for Africa
and so on.

I won't get my feelings hurt if this isn't welcomed input, but I just wanted to share my 2 cents.

[mcguirepr89]

Success :)

[mcguirepr89]

I've added some "minutes" to my geo-location, so the next detections should get plotted more accurately

[timsterc]

I just sniffed your changes and updated your station w/ the higher resolution, looks like NE Charlotte?

[mcguirepr89]

Yup 👍

[mcguirepr89]

Looking good

[DD4WH]

@timsterc
Tim, that really sounds really interesting!
Could you explain a little bit more, what the BirdWeather site is? I have not understood which hardware and software you are running and what the connection to BirdNET and BirdNETPi is. Also it would be very kind to explain what the purpose and the aim of this website is: https://app.birdweather.com/
On that website I can see stations all over the world, but I do not understand what they do at those sites and how I can access information on the sites. If I click on a site or an a species or on a soundscape, nothing happens at all.

Also, it would be good to have information on how you modified the Python script by Stefan Kahl and it would also be very kind of you to tell us where you published your modified code. Is it in your github account?

All the best ornithological wishes,

Frank DD4WH

[timsterc]

Hi Frank,

A fellow ham (I'm AB6CM) :-)

Apologies for doing a poor job of explaining what BirdWeather is and how it fits into the (growing) BirdNET ecosystem ... I'll try to do a better job here ... as my initial plan has morphed over the past month+.

I have a BirdNET-Pi installation running in my backyard in Moss Beach, CA (just south of San Francisco), which is also POSTing detections/soundscapes to the BirdWeather server. We're fortunate to have a nice variety of birds year-round visiting our backyard feeders, so you'll see pretty much continuous detections most of the day. I've not (yet) published my modifications to analyze.py, as I wanted to see what kind of response I got from Patrick and the BirdNET-Pi community. I'm 100% happy to share, just didn't want to cause any problems if people weren't really interested.

My vision of the BirdWeather site is to provide an accessible (central) site that accepts detections/soundscapes from a variety of sources, around the globe:
- BirdNET-Pi stations
- ~60 Live Streams / processed (live) on my Mac Mini M1, using a modified version of the BirdNET-Lite script from Stefan's repo
- Audio Live Streams (there is a network - most built around Raspberry Pi's - called Locusonus)
- YouTube Live Stream Cameras (a number of feeder cams, as well as various locations in nature preserves, etc)
- Drop-Off / Pop-Up Stations
- I have a little recorder called the Song Meter Micro (from Wildlife Acoustics) that I enjoy dropping off in different locations, leaving it for a few days, then running (yet another) BirdNET script which analyzes and posts those results (back-dated) to the server.
- Last, I'm working on a mobile app, which would basically be a live version of the BirdNET app, running the BirdNET AI on the mobile device, then posting results to the BirdWeather server. This would enable people who aren't Pi savvy, to purchase a Bluetooth microphone, and have a station in their backyard.

As I mentioned in my initial post, I had been contemplating building something along the lines of the Haikubox (albeit a slightly different approach), but after speaking with Stefan and understanding their (Cornell's) relationship with Haikubox, I have decided to pursue this as a pure passion project, with a view of getting people excited about the power of autonomous detections.

In terms of your usability questions ... of app.birdweather.com:

• The Top Species list on left is responsive to zoom, as well as the selected time period. If you zoom in on a particular area, even a single station, like mine in Moss Beach, you'll see the detection counts change.
• If you click on one of the Species, you'll see a Donut Plot appear below which shows detection counts (for that species) in a Time-Of-Day format.
• You can then click on the Soundscape, to scroll through the Spectrograms and listen to the Soundscape for that Species.

The app is still in it's early days ... and definitely hasn't been tuned for mobile devices ... so it's best viewed on a laptop with Chrome, Firefox, or Safari browsers.

Cheers!
-Tim

[mcguirepr89]

@timsterc -- I just wanted to highlight

I've not (yet) published my modifications to analyze.py, as I wanted to see what kind of response I got from Patrick and the BirdNET-Pi community. I'm 100% happy to share, just didn't want to cause any problems if people weren't really interested.

You'll need to go ahead and publish the new analyze.py to stay within the new BirdNET and/or BirdNET-Lite License, whichever of those models you are using for BirdWeather. Now that https://app.birdweather.com is public, so too must any of its source code which includes BirdNET or BirdNET-Lite material.

You can easily just make a fork of either my repo (since it looks like that is where you started), or of @kahst's and add your altered code and then add a link to that somewhere on https://app.birdweather.com. Once that's taken care of, we can proceed :)

[timsterc]

Perfect. I've forked BirdNET-Pi (https://github.com/timsterc/BirdNET-Pi) - just need to do a bit of clean-up, and I'll push the changes to analyze.py ... along with the other version I'm using for audio and YouTube streams ... and get a link added on the BirdWeather app.

Update - changes pushed to https://github.com/timsterc/BirdNET-Pi/blob/BirdWeather-Pi/analyze.py

[kahst]

Hey guys, thanks for your efforts. I really appreciate the work and thought you put into these projects. @timsterc we talked about this a couple of weeks ago, I really like the idea of birdweather and I would love to see it become an official BirdNET outreach tool. I would love to contribute an updated and customized model to the project - I always wanted such a web page but was never able to pull it off myself. Is there a way that we can:

• incorporate the newest prediction models (which are not public yet)?
• make it so that citizen scientists can easily join the project and set up their own listening station?
• people can click on a station and watch the scrolling spectrogram + detections?

What are your thoughts?

[mcguirepr89]

I would love to contribute an updated and customized model to the project

Would this mean BirdNET-Pi might get to use this, since its the "station" software?

[nilspupils]

Hi Stefan, thanks for joining the conversation!

• Is there a way that we can:

• incorporate the newest prediction models (which are not public yet)?

I think all of us would LOVE to see (and use) the newest model! Thanks for the great offer!

• make it so that citizen scientists can easily join the project and set up their own listening station?

This is already reality because of the tireless efforts of @mcguirepr89 et. al. !

• people can click on a station and watch the scrolling spectrogram + detections?

@mcguirepr89 already tried to integrate BirdNET-Demo with the scrolling spectrogram. There were problems related to Javascript and Python implementation which could not be resolved so far. See #24 for more info. We definitely need outside help here!

Best, Nils

[CaiusX]

Hi @timsterc

This looks like an exciting initiative and would be pleased if you included my INSTALLATION, which has been using @mcguirepr89 fantastic BirdNET-Pi installation running @kahst BirdNET-Lite unattended and unfailingly for over a month in South Africa with many power cuts, thunderstorms (listen to the Bearded Woodpecker samples) and now omicron!!

I think the BirdNET-Pi system is a great platform for distributed monitoring and your system is great for bringing all global installations into one spot, and providing the analytics and information links you are working on.

I'm sure @mcguirepr89 will be able to deploy new prediction models rapidly and, as a citizen scientist myself, found it easy to get the latest PRE-INSTALLED-IMAGE up and running with a Pi4, 64Gb card and cheap USB microphone (cant wait to get my hands on a PiZERO 2W and see what can be done on that)!!

I have put together rudimentary inforgraphics on a daily basis which I want to extend to

• include weather data from the net for the BirdNETPi site based on its location
• 7 Day, 30 Day and 365 Day aggregations
• Scrolling spectrogram + detections (please, please, please @kahst)!!!
• other visualization/analytic tools, as commented by @DD4WH

Shout what you need from me to get it onto BirdWeatherPi?

Best
African Penguin

[mcguirepr89]

Shout what you need from me to get it onto BirdWeatherPi?

Once Tim provides you with the station information, I will have a few edits to do to add STATION_NUMBER and STATION_NAME variables to birdnet.conf and then tweak birdnet_analysis.sh to check for the new variables for use as analyze.py arguments. I'll probably put this together as a testing branch that we can try out before pulling it into main.

[mcguirepr89]

PS My mother in law will be out of town for the first week of December, so I will be developing like crazy during that time -- until then, things are not so easy . . .

[timsterc]

@CaiusX - pleasure meeting you!

I've created a station - id is 61 - and the meta_data would be "BirdNET-Pi - Johannesburg, South Africa"

Cheers!
-Tim

[DD4WH]

@timsterc
Tim, Nice to see other HAMs also have other side interests not particularly connected to HAM radio ;-). I am working at the university of applied sciences HTW Dresden teaching students ornithological field methods, including the latest automatic ID possibilities (and their challenges), and I am really convinced that BirdNET is the best system available for that purpose at the moment (thanks to Stefan).
Thanks for your detailed information and thanks for making the modified BirdNET code available on github.
It seems yesterday my internet connection was not capable of showing what BirdWeather is meant to do: Today I was able to try things and see what it can show. Very cool and a very nice idea! I understand the automatic ID is done on your computer by processing audio streams that people have put into an internet stream, not locally on peoples devices, is that true?
Some of the visualisation that BirdWeather does, was very recently discussed here in the BirdNET-Pi context, eg. showing daily / weekly / monthly phenology of specific species and being able to click on a species and show spectrograms / listen to the recordings of that species. So, very nice to join efforts here ! :-)
I also really like the idea of Patrick running the ID locally and then sending the data to some kind of server (BirdWeather) to further process and aggregate data.

So, my ideal vision of the project would be:

• RPi running BirdNET-Pi locally identifying birds with BirdNET-Lite
• the ID should be running with standard parametres installed with a ready-to-go image that everyone can etch to an Sd card (like Patrick already implemented)
• on BirdNET-Pi there should be local graphs/statistics and running spectrogramm implemented
• advanced users should be able to configure parametres and other things for their specific purposes (running BirdNET-Pi stand-alone in the field etc.)
• all standard installations should send their data to BirdWeather
• BirdWeather should make it possible to aggregate data and make this available to users of the webpage
• BirdWeather should make it possible to aggregate data for further use in bird conservation and bird monitoring
• would be good to know about your plans on this last topic, Stefan

@kahst
Stefan, fantastic that you chime in and join the party :-).
As we already discussed some months ago when talking about running BirdNET-Lite on Windows (which we are still working on at the moment), it would be extremely helpful to be able to use the latest prediction model. In Central Europe, the ID is already very very good (and I have used BirdNET with several Terabytes of data), but in some species groups (owls for example) it could certainly be improved. Also some local ornithologists here are quite interested in NFC, but I do not know how far you have proceeded with that topic.
So, to answer your questions from my very personal point of view:

• incorporate the newest prediction models (which are not public yet)?
  YES, certainly we would be interested

• make it so that citizen scientists can easily join the project and set up their own listening station?
  As far as I tested the fantastic new image install procedure by Patrick @mcguirepr89 , this is already near-to-perfect from the software side on the RPi 4B (64bit OS). I am still testing alternative microphone, preamp and USB audio card setups, so there will be some choice of the mic set up according to peoples DIY-soldering-knowledge and financial budget

• people can click on a station and watch the scrolling spectrogram + detections?
  This has been on the wishlist since the very beginning of this project! So if you could help with implementing this, it would be really awesome!

This sounds really good to me and I am happy to help wherever I can!

All the best, Frank DD4WH

[timsterc]

Hi everyone!

Wow! What a wonderful bunch of responses to start the day with!

In terms of getting stations online, I'm happy to wait for @mcguirepr89 to get things into a testing branch, so that the BirdNET-Pi installation process remains super smooth for everyone!

@DD4WH - in answer to your question on the live processing of the (audio/YouTube) streams, yes, that's correct. I have the analyze.py script running on my Mac Mini M1 listening to just shy of 60 streams. I had to upgrade my Comcast/Xfinity plan to Unlimited, as my data usage has now skyrocketed! But it's been a super cool way to see how well @kahst BirdNET-Lite model works in a wide variety of soundscape conditions.

Working on another round of updates to the BW app, hopefully will be pushing an update later tonight.

... and @kahst in answer to your questions ... absolutely! I'm working on a separate page that would describe the different ways that a citizen scientist could get a station online, and logging data. Obviously including BirdNET-Pi!

Cheers!
-Tim

[mcguirepr89]

@timsterc I will do some installations with the new birdweather branch in the coming days (tomorrow is going to probably not be my most productive day, just a heads up in case I am not very responsive). So long as I can get that installing perfectly consistently, I'll put together a new Pre-installed image with BirdWeather pre-configured.
@CaiusX! I'm very eager to get your installation on BirdWeather as I know it to be a bountiful source of extractions. If it's alright with you, I'd like to be present either in the driver's seat or looking on during the shift so as to be able to catch things early. Also, as you know, I always learn a lot from your system :)

[timsterc]

@mcguirepr89 - I see you changed from UTC to locale timezone, which reminded me that I need to update the datetime formats to offset TZ form - as your detections are now POSTing in the future. I've made that change, and am thinking that instead of a fully open API that it would be better to go with a private (crypto) ID that I share with individuals. This would protect a station from getting spammed by either mistake or deliberately. That would also roll-up the station id and meta-data, so all I'd need to provide is the private ID. I should have that in place shortly.

[mcguirepr89]

That would also roll-up the station id and meta-data, so all I'd need to provide is the private URL. I should have that in place shortly.

analyze.py will no longer take the --s and --meta_data arguments?

[timsterc]

I'd be replacing those w/ just the --birdweather_id

Just thinking through potential ways that ill-intentioned folk out there could mess with the system, and a fully open POST API would be an easy target. Having a private URL should take care of it.

[mcguirepr89]

@timsterc One of the cool things about being in the US while there is BirdNET-Pi in South Africa, is watching the morning chorus of Karoo Thrushes and others between about 6pm-8pm your time. Check out the spectrogram view and watch the chatter on @CaiusX's installation -- you'll be green with envy -- I know I am! Detections usually start to trickle in for his installation right around now and usually get pretty dense. This is the 21st century birder's version of saying, "I saw a bunch of Karoo Thrush right over there!" only, the "right over there" is South Africa :)

[timsterc]

There is a YouTube live feeder cam that I put online a while back in SA, it's incredible!!! So cool!

[timsterc]

Ok, I've pushed the changes to analyze.py to my repo. I've replaced the --s and --meta-data with --birdweather_id

In order to keep those private, it's probably best for you to email me directly, and I'll provide the id's for both @mcguirepr89 and @CaiusX ... and create/assign any others as desired. My direct email is [email protected]

... there are also a couple additional modules I'm using to get the correct local timezone for the POST(s)

import pytz
from tzlocal import get_localzone

[timsterc]

Hi Guys,

Our internet is down :-( ... looks like since around midnight local time, Comcast are saying that they expect it back up around noon today, so I'll be offline most of that time (as will most of the stations as they originate here).

I'll post back when things are back up.

Update: back up as of an hour ago.

[CaiusX]

HI @timsterc and @mcguirepr89

Much activity while I was asleep -

Getting onto BirdWeather

@timsterc - will connect by e-mail & you can send details
@mcguirepr89 - I won't touch anything today besides my mucking about with info-graphics and weather. Feel free to pop in after your busy day (tvbolwyi) to get JHB onto Tim's platform, even if I'm not around, alternatively we can pick it up tomorrow & I can look over your shoulder.

Other Streaming Sites

@timsterc - I started a similar idea to yours a few months back but lack of coding skills & discovery of Patrick's work halted progress.

A rich stream source I found was AFRICAM - the free to view streams have really good sound and a variety of locations (Olifants, Tau, Tembe and Kwa-Maritane have pretty diverse locations and ecosystems). Maybe speak to them to see if you can/are allowed to include these streams on a public platform? I will forward contact details & my previous correspondence with Africam by e-mail.

Cutoff/Sensitivity

Optimal settings for settable parameters are essential to get as many true positives/true negatives. Analysis of running for a while and incorporation of an "exclude list" for identified false positives can be implemented as discussed previously [#45 ] - but this would be site specific and probably "unscientific" except for explicitly excluding publishing samples with, for example, human voices.

Identifying false negatives is a more difficult task - on my system certain pretty vocal species don't get recognized (Black Collared Barbet and Southern Boubou in particular). This is probably a function of the trained model and will undoubtedly be better as further BirdNET AI development takes place.

Best
Dennis (African Penguin in BirdNET-Lite)

[timsterc]

Here's an example (pulled from my admin dashboard) where @mcguirepr89's Pi heard a siren, and it incorrectly got a high confidence for a screech owl, but w/ the low probability, it pulls the score down ...

[CaiusX]

Hi Tim

Looks promising- in my case there are quite a few false positives which would be eliminated, and I'd be keen to explore whether it captures false negatives.

Could You provide insight into how the probability score is obtained?

[DD4WH]

Very nice to see your activities!

I would like to comment on the approaches to improve ID reliability.

• from my point of view the key to improving ID reliability is the exact definition of the ornithological goal of our specific projects:

• If our goal is to automatically gather as much data as possible for common species, the approach by Tim (and the approach by BirdNET [NOT BirdNET-Lite] to compare IDs with eBird distribution data is the right one. However, this is not an approach suited for the detection of rare/unexpected species

• If your goal is to detect rare and unexpected species, you should not use eBird data to calculate a new confidence score, but rely on the BirdNET-Lite confidence score alone. Maybe you can do some tricky post-processing things like taking into account the time of the day for owls for example --> here on my machine, I often get some false positive daytime owl IDs because of street noise / airplane noise.

@kahst
Stefan, what is your experience with post-processing BirdNET-Lite ID confidence scores? In your paper in MEE (https://besjournals.onlinelibrary.wiley.com/doi/pdf/10.1111/2041-210X.13571) you describe a post-processing by calculating a moving exponential average of the species-specific confidence score over three consecutive 3sec-chunks and it is stated that this is suited to lower the number of false positives. Would be nice to hear your experience with that.

In all, it would be good to more clearly define the goal of gathering the data (in an ornithological / nature conservation context maybe?), so based on this goal, we can think together how to improve IDs (which could mean increasing false negative number vs. false positives etc.). It all depends on what we would like to achieve :-).

For example, I am specifically interested in identifying Eurasian Pygmy Owl in my region. This species (and many others) would probably go undetected when eBird data would be used in postprocessing. However, I fully acknowledge that using the eBird data approach would significantly improve IDs when the goal is to gather immense amounts of records for common species . . . it all depends and our mileages may vary :-).

[mcguirepr89]

Hi @timsterc -- I pulled the latest and am testing now -- looks like a successful POST, so that's great!

Soundscape POST Response Status -  201
{ "timestamp": "2021-12-01T12:17:57-05:00","lat": 35.3112,"lon": -80.749,"soundscapeId": 861375,"soundscapeStartTime": 6.0,"soundscapeEndTime": 9.0,"commonName": "Mallard","scientificName": "Anas platyrhynchos","algorithm": "alpha","confidence": 0.70852125 }
Detection POST Response Status -  201

https://app.birdweather.com, however, did an interesting thing. It POSTed the Soundscape as occurring in the 12pm hour (which is correct), but the timestamp on the spectrogram excerpt is 3 hours behind, using 9:17am as the time.

[timsterc]

Hmmm ... I'll dig into that.

Yep, detection's are getting properly stamped, but need to pull those into the soundscape. Working on a few other things too, so probably get this in the next update [probably tomorrow]

[mcguirepr89]

Spectrogram time-stamps look good now 👍

[timsterc]

Got that (and a few other bugs) fixed up, as well as adding the ability to click on a station or one of the notifications, which will then zoom to that location.

[mcguirepr89]

Very cool -- I love these improvements!

[mcguirepr89]

Hi @timsterc -- we got @CaiusX updated and POSTing to BirdWeather 👍

[timsterc]

@mcguirepr89 - I'm up and running w/ the new pre-installed image now as well!

[mcguirepr89]

Hi, @timsterc -- just wanted to drop you a line to say that I have been noticing your great work polishing BirdWeather up. It gets better every day :) 🚀

[Phaeton]

I read the entire discussion and it’s not entirely clear to me. Is the entire stream sent to Birdweather or just the extracted clips from the local BirdNET-Pi?

[timsterc]

Patrick is correct - best to be specific. If there are any detections made within the audio file (I believe the default recording length is 15sec) then the entire file is uploaded to the BirdWeather server, along with the detected species information. When I originally answered I was comparing this to the live streaming stations (YouTube and Audio) that are also supported on BirdWeather as well.

[nilspupils]

What is the best way to reduce upload volume. My second system has literally thousands of detections a day, and it sucked up my mobile data volume quite fast, so that i had to take it offline for now. Would be great to put it back online soon!
Best, Nils

[timsterc]

@nilspupils - I guess that's a good problem to have - never too many birds :-)
I'll have a think about what we might do on our side - possibly allowing detections to be posted w/o a soundscape.
Without any code modifications, the only thing I can think of would be to increase the Minimum Confidence, which would then cut back on the number of detections/day - which I'm guessing isn't desirable. Maybe it would make sense to have separate Minimum Confidence for BirdWeather Posts?

[nilspupils]

If i limit the number of recordings in extract_new_birdsounds.sh to 20 per species per day this would probably reduce the upload volume?

[Phaeton]

Is it feasible to move the Birdweather POST logic from the analysis.py/server.py scripts to extract_new_birdsounds.sh? Is this a feature request that could gain some traction? I had noticed the full clip going earlier and have now modified my recording length to be much shorter.

[timsterc]

Figured I'd post in this thread as well ...

The goal with BirdWeather is to provide both live visualization, but also to be a fully open databank of all the detections uploaded. We're using GraphQL - so for those familiar with it:

• API - https://app.birdweather.com/api/index.html
• Online GraphQL Query Tool - https://app.birdweather.com/graphiql

We've also finished the first pass at our data explorer - https://app.birdweather.com/data - which also supports exporting results via CSV.

Cheers!
-Tim

[bvhillier]

Hi Tim, really great project. I cannot see the CSV export option on the Data Explorer page, what am I missing? cheers, Ben

[timsterc]

@bvhillier - you'll need to create an account - https://app.birdweather.com/login - then you'll be able to export.

[bvhillier]

Thanks for quick response, i created account and now see the export tab but I after specifying my email address the detection file is not delivered, not in spam either. I tested with a small list of 5 detections (cuckoo, UK, June 1st only, >90% confidence) but still nothing. Any thoughts? cheers..

[manno-xx]

I am thinking of setting up a birdnetpi station and linking it to birdweather but was wondering about what happens with the data. I’m frequently posting observations to https://waarneming.nl/ which is similar/same to https://observation.org/ and hope the data is being put to some good use. Is there a links from birdweather to a “standard” observation platform? Other than a cool visualization, i’d like to contribute to a better understanding of what goes on with birds (in this case) in our world :)

[timsterc]

@manno-xx - we're not actively pushing observations to any other sites yet. We do have an open API, and we're starting to see others using that now. It's still early days ... I've not yet done any real promotion of the site ... so it's basically just word of mouth at this point, but my hope is that this becomes a valuable resource for researchers! One of the main reasons I have chosen to focus so heavily on the visualization side (up front) is to make the experience something where people would want to keep their stations online. Being able to share with family/friends, etc. I'm really pleased with the continuity of the BirdNET-Pi stations - it's been very cool to see them grow, and continue to stay online!

I just checked both sites - I'm not seeing any (obvious) API references for either site? Do you know if they support posting observations other than using their app as an end user?

[manno-xx]

@timsterc Don't get me wrong, I love the idea and an easy to browse and visual attractive view on the data is already valuable!
Where can I see the API end points? I suppose it's in the BirdNET-Pi code base as that is the source of the data?

I hope in some near future I'll find the time to set up a BirdNET-Pi station as well. (moved to the, sort of, great outdoors for as far as that is possible in the Netherlands, but still working on the renovation and electricity and wifi coverage stuff :) )

To be honest, I'm not aware of an API for mentioned websites. The NIA (neural network I suppose) the websites use is maintained by Naturalis, A well known museum (with a virtual version) that also does research and education and with links to the Leiden uni iirc (some info on the API). It is an API, but not open I suppose and only doing visual identification(?) and I think only in support of the websites, those apps and their data set...
Having had some tangential connections to people there, I think I know they are open to innovative ideas and visualisations as teaching kids (playfull etc.) is one of their goals for sure.

[cam-el]

Hi Just wanted to say what a fantastic addition this is to an already amazing project. I can check my backyard activity from work, Thanks!

Can I suggest an option for metric figures :)

[pddpauw]

this is a great project, thank you so much! Useful to see own and other's bird's activities.
Couple of suggestions:
i) language selection
ii) improved view for mobile sites: seeing the species overview per station is not possible in mobile view
iii) link with other observations (e.g. observations.org) to have an even more holistic view on observations

Keep up the great work!
Pieter

[programmdesign]

I'd also appreciate if the mobile version would allow me to see the "Top species" list on my mobile phone.

[jpasqua]

This looks really valuable for a project I am doing with students for a local Audubon chapter. Has anyone looked at this for intermittently connected devices? In particular, we will have multiple Birdnet-Pi devices in the field with no internet access. There are volunteers on site periodically who could create a hotspot on their phone to allow for data upload to birdweather. It appears that currently uploads to birdweather are done during each detection cycle; which will fail most of the time in this setup.

I'm thinking that maybe we'll have a process in the background that notices when wifi is available, then upload backlogged data to birdweather from the "processed" directory. It would remember the last timestamp it uploaded and pick up from there upon the next connection.

[mikekeith]

i suspect not, but.. :) is there anyway to remove or edit recordings/detections that my station has submitted? mainly thinking about privacy issues, and incorrect detections (who knew that me coughing sounded like a ring necked pheasant :) )

if not via the web frontend, can it be done via the api?

[pddpauw]

you can remove those recordings from the Birdnet Pi page (select the recording and press the bin icon)

[mikekeith]

ahh interesting! and that removes them from birdweather? i had no idea :)

[waynecarnold]

No I do not think that is how it works. I think he misunderstood and thought you wanted to remove only from your device. Wayne

…

On Tue, Oct 31, 2023 at 4:25 PM mikekeith ***@***.***> wrote: ahh interesting! and that removes them from birdweather? i had no idea :) — Reply to this email directly, view it on GitHub <#82 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BAPGSXH7J4FUXMEOZ7BB7TDYCFNCXAVCNFSM5I4UOZI2U5DIOJSWCZC7NNSXTOKENFZWG5LTONUW63SDN5WW2ZLOOQ5TONBTHE2DGMQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.*** com>

[bvhillier]

Brilliant, thanks! Ben Ben Hillier MRes PhD FGS Regional Representative North London, British Trust for Ornithology (BTO) www.bto.org This message and any accompanying attachments are intended for the addressee named and may contain confidential or privileged information. If you are not the intended recipient do not read, use, distribute or copy this message or attachments. Please notify the sender immediately and delete the message. From: Tim Clark ***@***.***> Reply to: mcguirepr89/BirdNET-Pi ***@***.***> Date: Wednesday, 20 December 2023 at 16:52 To: mcguirepr89/BirdNET-Pi ***@***.***> Cc: Dr Ben Hillier ***@***.***>, Mention ***@***.***> Subject: Re: [mcguirepr89/BirdNET-Pi] BirdWeather Site (Discussion #82) @bvhillier - you'll need to create an account - https://app.birdweather.com/login - then you'll be able to export. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: ***@***.***>

[RFischer2024]

I received my BirdWeather PUC a few days ago. Setup was easy and it was performing perfectly for two days. Now, it won't stay online. Has anyone else experienced a similar issue? There are new batteries in the unit. I can get it to connect, and go online, but it then goes offline a few minutes after. Internet connection is strong. I cannot find a tech support link on their website.

[timsterc]

On the top of the Quick Start Guide - there should be text that says email ***@***.*** - are you not seeing that? https://birdweather.com/quick-start-guide

…

-Tim

On Sat, Dec 23, 2023 at 12:05 PM RFischer2024 ***@***.***> wrote: I received my BirdWeather PUC a few days ago. Setup was easy and it was performing perfectly for two days. Now, it won't stay online. Has anyone else experienced a similar issue? There are new batteries in the unit. I can get it to connect, and go online, but it then goes offline a few minutes after. Internet connection is strong. I cannot find a tech support link on their website. — Reply to this email directly, view it on GitHub <#82 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAI2KUZZK5HI7YRITXRF7Z3YK42OZAVCNFSM5I4UOZI2U5DIOJSWCZC7NNSXTOKENFZWG5LTONUW63SDN5WW2ZLOOQ5TOOJTGU2DANA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[RFischer2024]

I see that now. Thank you!

[bvhillier]

My PUC reads 5 degrees celsius higher than actual temperature (i've calibrated with other thermometers) so can this be adjusted somehow? Not a priority just enquiring.

Everything else is good - it's a nice device.

Some feedback - only stays LIVE with wifi and USB charging. Even on fully charged batteries and strong wifi it drops to offline. Worked fine offline with batteries on a day trip although upload of 12 hours worth of detections (+/-3000) took more than 48 hours.

[AndrewCauldwell]

@bvhillier, my PUC also overstated the temp and pressure readings were unrealistic. But had a power outage which caused device to restart and it has corrected itself.
There is a reboot function in the menu which likely achieves same result.

[bvhillier]

Belated thanks Andrew!, good to know yours is getting more reliable readings now. I've tried that with no success but as a I said its not a priority, cheers, Ben

[ON5MF]

Hi, can anyone tell me what the 'Automatic Gain' does in the settings on the birdweather site?
I have been looking around for a while but can't find any explanations.

[bvhillier]

I am interested to replicate some research I undertook using bioacoustic data from two seasons across all of Norway. Is it possible to know how many hours per day each PUC device was operational and recording? This is important information when analysing the number of detections across a set of acoustic devices. I downloaded some data but this information did not seem to be in then standard csv format, thanks, Ben

[timsterc]

Hi Ben, Best to reach out to us directly - ***@***.*** - for things related to the PUC. Cheers! -Tim Founder - BirdWeather

…

On Mon, Sep 9, 2024 at 9:52 AM Dr Ben Hillier ***@***.***> wrote: I am interested to replicate some research I undertook using bioacoustic data from two seasons across all of Norway. Is it possible to know how many hours per day each PUC device was operational and recording? This is important information when analysing the number of detections across a set of acoustic devices. I downloaded some data but this information did not seem to be in then standard csv format, thanks, Ben — Reply to this email directly, view it on GitHub <#82 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAI2KU5B4ALRPIYMZG2S5ULZVXG5NAVCNFSM5I4UOZI2U5DIOJSWCZC7NNSXTOSENFZWG5LTONUW63SDN5WW2ZLOOQ5TCMBVHEZTOMBT> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[bvhillier]

Ah OK, will do!