Feature Request - Geo-biased Analysis & eBird Integration

[SterlingMcCall]

1. Would it be possible to specify a region/location in order to geographically bias the species detection? This would help avoid false positives for unlikely species. For example, I had a sequence of a white-throated sparrow photos and while many were detected correctly, some shots from the back were detected as Brewer's sparrows, which could have been eliminated from consideration given the fact that the images were taken in Wisconsin and not the western US. My images do not contain coordinates within the metadata, so this would need to be a manual option when setting up the analysis.
2. If the user could link their eBird account, would it be possible to use the API to pull the user's checklists and GPS tracks, then associate them automatically with the photo timestamps? This would have the added benefit of being able to write the coordinates to the xmp file while also assisting the automatic species detection.
3. It would be really cool if you could then suggest photos to upload to the eBird checklists, automatically suggesting the appropriate species.

I realize all of this would be considerable work, but I think it would be a spectacular speed-up to the whole flow for those that have interest in contributing data to eBird.

[SanjaySoniLV]

Hey! Thanks for the input. Regarding improving bird ID by geographically biasing the classifier... This is a good idea but I'm afraid it's just going to be too difficult to implement for the time being. I would rather add more region specific models (ex. A European model) before getting into within-region geography. Regarding ebird integration... I had this idea myself. I'm open to it, I think it could be pretty cool to link a checklist in a way, for a number of purposes including search. I'll experiment with this. It seems like the user would need to give Kestrel their ebird API key, since Cornell doesn't have any system to connect 3rd party apps to ebird. And as far as uploading to eBird/Malaculay... There's no programmatic way to interface with their platform, so the best I could do was offer a "quick export" feature that let's you bypass opening a full editor if your goal is to just do light tweaks and export a photo for contribution purposes. What do you think about this? Maybe it could save to a "Kestrel Exports" folder, and the files can be named by the species tag of the scene.

…

________________________________ From: SterlingMcCall ***@***.***> Sent: Sunday, April 5, 2026 11:47:12 AM To: SanjaySoniLV/ProjectKestrel ***@***.***> Cc: Subscribed ***@***.***> Subject: [SanjaySoniLV/ProjectKestrel] Feature Request - Geo-biased Analysis & eBird Integration (Issue #27) [https://avatars.githubusercontent.com/u/8882589?s=20&v=4]SterlingMcCall created an issue (SanjaySoniLV/ProjectKestrel#27)<#27> 1. Would it be possible to specify a region/location in order to geographically bias the species detection? This would help avoid false positives for unlikely species. For example, I had a sequence of a white-throated sparrow photos and while many were detected correctly, some shots from the back were detected as Brewer's sparrows, which could have been eliminated from consideration given the fact that the images were taken in Wisconsin and not the western US. My images do not contain coordinates within the metadata, so this would need to be a manual option when setting up the analysis. 2. If the user could link their eBird account, would it be possible to use the API to pull the user's checklists and GPS tracks, then associate them automatically with the photo timestamps? This would have the added benefit of being able to write the coordinates to the xmp file while also assisting the automatic species detection. 3. It would be really cool if you could then suggest photos to upload to the eBird checklists, automatically suggesting the appropriate species. I realize all of this would be considerable work, but I think it would be a spectacular speed-up to the whole flow for those that have interest in contributing data to eBird. — Reply to this email directly, view it on GitHub<#27>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AU5VT4GM2PHBJ3VAQK6ISP34UKEZBAVCNFSM6AAAAACXNLXXJ6VHI2DSMVQWIX3LMV43ASLTON2WKOZUGIYDQMBXGEZTSNA>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[SterlingMcCall]

That's a fair point, prioritizing more regions makes more sense than subregions.

I looked into the eBird API a bit, and it seems that it does not offer a nice way to get one's own checklists and geo-tracks. You can pull public checklists by ID, but I don't see a way to do it in bulk for a given user. Additionally, the checklist track data is not exposed via the API.

I did use Gemini to write a python script that will use the website endpoints with your username and password (rather than an API key) to download checklists as CSV (although that could be improved to download the JSON format from the proper API). It also downloads the track as a KML file, but unfortunately the track data is missing timestamps with the coordinates, so there would be no way to accurately map the coordinates back to specific images.

ebird_checklist_downloader.py

Usage: python ./ebird_checklist_downloader.py <username> <password> [checklist_id]

I usually track my bird-photography hikes/walks with my Garmin watch as well, so I might just make my own script to pull the gpx files and leverage exiftool to write the coordinates.

---

Regarding Macaulay/eBird Photo Uploads

I think that could be a nice-to-have for some people. But personally, basic support for color labels and arbitrary tags (not just species/family) would be more helpful, as I could just slap an "eBird" tag on the ones I want to upload, then filter and export them from whatever editor I prefer.