diff --git a/Help/bugs.html b/Help/bugs.html index 97eb56af..f05b9aa2 100644 --- a/Help/bugs.html +++ b/Help/bugs.html @@ -8,12 +8,12 @@

Have you hit on a problem, or perhaps you have a suggestion?

-If you have, then I'd be happy to hear about it. I keep track of issues and suggestions on Github. +If you have, then I'd be happy to hear about it. I keep track of issues and suggestions on Github.

Take a look at what others have been reporting, if you think you have a similar problem or suggestion, add a comment to the thread.

-

If, on the other hand, you think you've got something new, go on and add an issue - I'll get back to you as soon as I can.

-I can't promise I'll make everything right, but I probably won't fix what I don't know about ¯\\_(ツ)_/¯

Meanwhile, I hope you find Chirpity useful!

+

If, on the other hand, you think you've got something new, go on and report an issue or start a discussion - I'll get back to you as soon as I can.

+I can't promise I'll make everything right, but I probably won't fix what I don't know about ¯\\_(ツ)_/¯

Meanwhile, I hope you find Chirpity useful, and happy birding!

\ No newline at end of file diff --git a/Help/settings.html b/Help/settings.html index f2bff590..c52eabcc 100644 --- a/Help/settings.html +++ b/Help/settings.html @@ -37,35 +37,6 @@ of the files and the location that is set. - -
Experimental Features
- - - Context Mode - In this mode, Chirpity will use the surrounding context when making predictions. This helps mitigate - against false positive detections. - - - SNR filter - The SNR (Signal to Noise Ratio) filter can only be enabled when using the CPU backend. When enabled, Chirpity will disregard audio segments with no distinct sound event. - A stronger signal is required for a prediction to be attempted when higher signal-to-noise values are set. - The purpose is to deliver significant speed gains, however, the setting may cause Chirpity to miss quieter, more distant calls. - - - - High Pass filter - This applies a filter to the audio, removing the sound below the frequency selected. - Intended primarily for recordings made in in urban settings, this is desirable to compensate for strong low-frequency noise which is not present in the training data. - - - - Low Shelf filter - This applies an audio filter similar to the High Pass filter, however, rather than simply remove noise, it reduces the volume of low frequency noise. - You can set the threshold frequency, and the strength of the filter to apply (Attenuation). - Also intended for recordings made in in urban settings, this helps tune the audio to the noise profile of the training data. - - -
Location
@@ -128,7 +99,7 @@ Chirpity will use the native Chirpity model for predictions.
  • - BirdNet will use the BirdNet model. + BirdNET will use the BirdNET model developed by Stefan Kahl et al. from the K. Lisa Yang Center for Conservation Bioacoustics
    • @@ -146,6 +117,34 @@ Larger values lead to faster processing, especially on long files. + +
    Experimental Features
    + + + Context Mode + In this mode, Chirpity will use the surrounding context when making predictions. This helps mitigate + against false positive detections. + + + SNR filter + The SNR (Signal to Noise Ratio) filter can only be enabled when using the CPU backend. When enabled, Chirpity will disregard audio segments with no distinct sound event. + A stronger signal is required for a prediction to be attempted when higher signal-to-noise values are set. + The purpose is to deliver significant speed gains, however, the setting may cause Chirpity to miss quieter, more distant calls. + + + + High Pass filter + This applies a filter to the audio, removing the sound below the frequency selected. + Intended primarily for recordings made in in urban settings, this is desirable to compensate for strong low-frequency noise which is not present in the training data. + + + + Low Shelf filter + This applies an audio filter similar to the High Pass filter, however, rather than simply remove noise, it reduces the volume of low frequency noise. + You can set the threshold frequency, and the strength of the filter to apply (Attenuation). + Also intended for recordings made in in urban settings, this helps tune the audio to the noise profile of the training data. + + diff --git a/js/BirdNet2.4.js b/js/BirdNet2.4.js index 56d5b6a2..d019b399 100644 --- a/js/BirdNet2.4.js +++ b/js/BirdNet2.4.js @@ -368,6 +368,7 @@ class Model { const tb = paddedTensorBatch || maskedTensorBatch || TensorBatch; const prediction = this.model.predict(tb, { batchSize: this.batchSize }) + let newPrediction; if (this.selection) { newPrediction = tf.max(prediction, 0, true); diff --git a/package.json b/package.json index cbbfd3e8..583b0d0a 100644 --- a/package.json +++ b/package.json @@ -96,7 +96,7 @@ ], "mac": { "target": { - "target": "pkg", + "target": "dmg", "arch": "arm64" }, "files": [