6
$\begingroup$

My first post here...

Using BirdNET, I have created two selection tables, each at a different Confidence setting, for display in RavenPro. The source files are 48k, 2-channel, 16-bit signed .WAV.

I need to simultaneously compare those two selection tables in RavenPro (1.6.5 beta) on the same screen. I can simultaneously display the same sound file (one above the other) via View > New > Spectrogram View, as shown in the screenshot below. However, whenever I load the selection tables, both tables always display together in both spectrograms, superpimposed over each other. I want to separate the two tables, one in each spectogram.

Question:

How to simultaneously view two different selection tables, each in its own spectogram?

Windows 10

enter image description here

$\endgroup$

2 Answers 2

1
$\begingroup$

My understanding is that selection tables are related to data (time series/spectrogram).

So, the easiest way is to open two sound files, then under "Windows" tab > "Tile windows vertically"

and you get something like this: enter image description here

each of the two windows can have its own selection table

Note: I would find it difficult to have one timeseries/spectrogram with two or more selection tables (if that is was you were looking for, but from OP I conclude this is not the case), as you will clutter the timeseries/spectrogram with boxes and confusing colours

$\endgroup$
1
  • 1
    $\begingroup$ I have used a similar approach to this to simultaneously scan through the same dataset but with two different decimation rates (to identify calls that spanned from very low, 10s of Hz up to about 7kHz, and sometimes you only got one part of the call). It's a bit tricky but did the job and think this is a reasonable approach! The full paper (Nieukirk et al. 2016 JASA 140, EL274-EL279) is here: repository.library.noaa.gov/view/noaa/14697 $\endgroup$
    – selene
    Commented Apr 27, 2023 at 4:48
0
$\begingroup$

Good question! I will discuss several options.

  1. To achieve what you are looking for -- two spectrograms, each with a different selection table -- you will want to open two sound windows so you can open each selection table into a different sound window. Raven will position the two sound windows nicely if you select "Tile Windows Vertically" or click on the "Tile Windows Vertically" button on the Window toolbar. However, it is difficult to navigate through the recording because the two sound windows are not linked in any way. Fortunately, you can link the two spectrograms by Time Position, Time Scale, Frequency Position, Frequency Scale and perhaps a few other properties, as described in Chapter 3 of the Raven Pro User's Manual, which you can open from the Help menu. This will allow you to easily move and zoom the two spectrograms together on the time and frequency scales. In the screenshot below the selection table on the top was created with a 0.01 threshold and the one below with a 0.25 threshold.

2 linked sound windows in Raven Pro

  1. I suspect before too long you will find that the above protocol is tedious and inefficient. Another strategy that is a bit less work for your hands and your eyes would be to open both selection tables into the same sound window and toggle the "Draw" control to hide and show one of the selection tables. In the demo below the selection table created with an 0.25 threshold is displayed with light aqua boxes and the selection table with the 0.01 threshold is displayed with dark blue boxes.

toggle selection table display off and on

  1. Both the strategies above are labor intensive enough that they will limit how many selections you can review. They also do not lend themselves to quantitative analysis. For a slightly more rigorous analysis you could tabulate and plot precision by threshold, precision being the number of selections that correctly identified the intended bird species divided by the total number of selections in the BirdNET output. One way to accomplish this would be to annotate all the selections created by BirdNET to indicate whether they are TP ("true positives" = correctly identified the species) or FP ("false positives" = did not correctly identify the species). The Selection Review feature in Raven Pro is an efficient way of doing this. You could extend this idea by tabulating and plotting Precision x Recall across a range of thresholds.

Raven's selection review feature

I admit I have gone pretty far afield of the OP question so will stop here. I see a Bioacoustics Stack Exchange post on Software options for diagnosing the performance of a detection routine that is germane to point 3 above. If you get stuck trying any of the Raven suggestions above, [email protected] is quite responsive.

$\endgroup$
2
  • $\begingroup$ Thanks for confirming (implicit) my answer and for adding more options $\endgroup$
    – WMXZ
    Commented Apr 29, 2023 at 12:53
  • $\begingroup$ Do you have advice about the best etiquette for agreeing with and elaborating on someone else's answer? $\endgroup$ Commented Apr 30, 2023 at 13:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.