Fresh BirdNET installation detecting birds incorrectly (e.g. from a remote geography) #956
-
Hi, I've freshly installed BirdNET yesterday on a freshly-installed Raspbian Bullseye RPI4 system. The system is detecting birds incorrectly - e.g. this morning it found a Blue-headed Parrot, even though I am in western Washington state. I've run species.py and then rebooted to see if that might nudge the system to look at local species, but no luck. If it's useful at all for debugging this, the most common species detected is a Yellow-bellied Elaenia, which does not range further north than Mexico. Thanks for any suggestions on what I might try here! |
Beta Was this translation helpful? Give feedback.
Replies: 10 comments 11 replies
-
Could you try running this command and seeing if things come back to normal?
|
Beta Was this translation helpful? Give feedback.
-
Thanks - unfortunately this doesn't seem to help. When I installed, I created a local version of newinstaller.sh (from |
Beta Was this translation helpful? Give feedback.
-
Done. Doesn't seem to have made a difference, other than it's now detecting a "White-browed piculet" (native to Asia) with 75%-95% confidence. Given that I am in Washington state, doesn't seem likely this is accurate. :) Happy to give you ssh access to the RPI4 through a Tailscale VPN if you want to poke around - it's a fresh install of Bullseye so nothing sensitive there. |
Beta Was this translation helpful? Give feedback.
-
Awesome - thank you! Super appreciate you fixing it up for me. We just
moved to a very rural area with all sorts of interesting birds to track, so
BirdNET will be super useful to us.
…On Tue, Jun 13, 2023 at 4:22 PM ehpersonal38 ***@***.***> wrote:
Okay, got your system all fixed up! Turns out you were using the old model
so the labels file was mismatched. I just changed the model to the new one
and all is well. You also had some weird issue after that where the
analysis was hanging... I still don't know why it happened, but a few
restarts of services and it seems to be chugging away again. Let me know if
you need any other assistance!
—
Reply to this email directly, view it on GitHub
<#956 (reply in thread)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAEF7WSQTANJRIALSWRYLATXLDY47ANCNFSM6AAAAAAZDXXDEM>
.
You are receiving this because you authored the thread.Message ID:
***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
(apologies - solved this one myself by switching to the BirdNET_GLOBAL_6K_V2.4_Model_FP16 (2023) database via the Tools menu. Should that be the default these days ???) ~~ Me too. Also near Seattle/Tacoma. Merlin on the iPhone is correctly identifying the white-crowned sparrow that I'm sitting here looking at, but BirdNET on the pi.is incorrectly identifying it as a Malagasy Bulbul or a Long-tailed Koel, neither or which are from anywhere near here. This is on a clean pi4 using latest raspi os 64 bit. ~~ |
Beta Was this translation helpful? Give feedback.
-
I've had the same problem, and hopefully solved it by switching to the new model. Is there any way to get a list of corresponding names for the new model. E.g. what should a blue headed parrot really be? I'd like to have a go at updating the detections table with the correct names and renaming all those mp3 files to match. I have over a week of data which it would be a shame to lose. Did this data get uploaded somewhere? Is bad data percolating up? Thanks, Kevin |
Beta Was this translation helpful? Give feedback.
-
Read the thread. Switch which db you use.
…On Tue, Aug 22, 2023 at 2:19 PM rovermeyer ***@***.***> wrote:
What is the fix for this? I did a new install yesterday. I live in
Maryland and all of the birds I'm detecting are from Africa.
—
Reply to this email directly, view it on GitHub
<#956 (reply in thread)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAETMKXPCHCFVPI7NAWPB33XWUO6BANCNFSM6AAAAAAZDXXDEM>
.
You are receiving this because you commented.Message ID:
***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
I've seen an issue fairly frequently where incomplete recordings cause analysis to hang, but I've never managed to capture a good example for debugging! |
Beta Was this translation helpful? Give feedback.
-
I had the same problem. Switching the model fixed it, thanks. |
Beta Was this translation helpful? Give feedback.
-
I had the same issue. Seems like the default model should be changed if people keep having this issue? |
Beta Was this translation helpful? Give feedback.
Okay, got your system all fixed up! Turns out you were using the old model so the labels file was mismatched. I just changed the model to the new one and all is well. You also had some weird issue after that where the analysis was hanging... I still don't know why it happened, but a few restarts of services and it seems to be chugging away again. Let me know if you need any other assistance!