Reporting multiple detections as one?

[Svardsten53]

Would it be possible to report several detections for a species within the same 15 seconds interval as just one detection? Frequent birds tend to otherwise fill the "Today's Detections" list. See example below.

The disadvantage of only reporting one hit may be that, for example, you would not be aware that several birds would fly in the sky (migratory birds). However, one can never be sure if it is sounds from several individuals or just one that has sounded several times.

This is just an idea from me, others probably have other views.

[mcguirepr89]

Hello, my friend! Great question

Would it be possible to report several detections for a species within the same 15 seconds interval as just one detection? Frequent birds tend to otherwise fill the "Today's Detections" list. See example below.

While I'm sure this would be possible with edits to analyze.py, I won't be implementing anything like that for BirdNET-Pi since its 24/7 monitoring is designed to take into account everything that can be detected. And while I understand your desire to have a "Today's Detections" will fewer redundancies, the database itself exists to keep these extensive records.

I will put together a script for you, however, that will cull the database of its duplicate entries. Once you've been able to test the script to be sure it does what you'd like, we can set it up to cull the database of duplicates at regular intervals so that your "Today's Detections" view can more closely resemble what you have in mind.

I will get back to you when the script is ready to test and we'll go from there ~~

My best regards as always,
Patrick

[Molokai]

Oh and if you want a new table with deduplicated detections, you could create it like so:

CREATE TABLE deduplicated_detections
SELECT Date, Time, Sci_Name, Com_Name, MAX(confidence) AS Confidence, Lat, Lon, Cutoff, Week, Sens, Overlap, COUNT(*) AS Count
FROM detections
GROUP BY date, time, sci_name, com_name, Lat, Lon, Cutoff, Week, Sens, Overlap;

This adds the column Count that indicates how many detections were "collapsed" into one. This assumes that Lat, Lon, cutoff, Week and Sens fields are identical for all the entries.

I would like to come up with a query that can be run periodically to add new deduplicated entries. If there was a combined date/time field I could just have done this:

INSERT INTO deduplicated_detections
SELECT Date, Time, Sci_Name, Com_Name, MAX(confidence) AS Confidence, Lat, Lon, Cutoff, Week, Sens, Overlap, COUNT(*) AS Count
FROM detections
WHERE DateTime > (SELECT MAX(DateTime) FROM deduplicated_detections)
GROUP BY date, time, sci_name, com_name, Lat, Lon, Cutoff, Week, Sens, Overlap;

It is probably possible to combine the date and time fields in order to do something similar. I'll take a closer look when I have more time.

Ah this was really quite easy to solve!

INSERT INTO deduplicated_detections
SELECT Date, Time, Sci_Name, Com_Name, MAX(confidence) AS Confidence, Lat, Lon, Cutoff, Week, Sens, Overlap, COUNT(*) AS Count
FROM detections
WHERE TIMESTAMP(date,time) > (SELECT MAX(TIMESTAMP(date,time)) FROM deduplicated_detections)
GROUP BY date, time, sci_name, com_name, Lat, Lon, Cutoff, Week, Sens, Overlap;

[mcguirepr89]

@Molokai This is awesome!! You've given me a lot to work with. Thank you so so much and I will probably be following up with lots of questions :)

My best regards,
Patrick

[Molokai]

Fire away! Happy to help 👍

[mcguirepr89]

I've just pushed the change to viewdb.php that now uses:

SELECT Date, Time, Sci_Name, Com_Name, MAX(Confidence) 
FROM detections 
WHERE Date = CURDATE() 
GROUP BY Date, Time, Sci_Name, Com_Name 
ORDER BY Time DESC;

to just show the highest confidence and "Report Multiple Detections As One"

Thank you again, @Molokai!!!

@Svardsten53 The community solved this for us!! :) :) :)

[Svardsten53]

Super! I have included the new code in my own customization and can see the difference.