Skip to content

Commit

Permalink
Completed worker refactor. Now listmodel handles all list updates and…
Browse files Browse the repository at this point in the history 
… species logic.

The outpist of the list worker is cached

Removed all mdata_model functionality from predict model workers
Aligned BirdNet / Chirpity species names

Added a glob to exclude *.md files from package  = c. 5MB of unnecessary content
optimised the birds/migrants filter function: 57ms --> 2ms

Updated Settings help.
  • Loading branch information
@Mattk70
Mattk70 committed Feb 7, 2024
1 parent 0cdab07 commit 6567968
Show file tree
Hide file tree
Showing 7 changed files with 139 additions and 338 deletions.
122 changes: 76 additions & 46 deletions Help/settings.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -10,24 +10,39 @@
<tr>
<td colspan="2" class="text-center text-bg-light"><h5>Predictions</h5></td>
</tr>
<th>Model</th>
<td>Choose the model to use for prediction.
<ul>
<li>
<b>Chirpity</b> will use the native Chirpity model for predictions.
</li>
<li>
<b>BirdNET</b> will use the <a href="https://github.com/kahst/BirdNET-Analyzer" target="_blank">BirdNET</a> model developed by Stefan Kahl et al. from the K. Lisa Yang Center for Conservation Bioacoustics
</li>
</ul>
</td>
</tr>
<tr>
<th>Set Threshold</th>
<th>Confidence Threshold</th>
<td>This setting determines the minimum confidence level required for the model to report a detection.
Increasing the value reduces false positives but increases the risk of missing genuine calls.
A value around 50% provides a good balance between detecting genuine calls and minimizing false positives.
</td>
</tr>
<tr>
<th>Search for</th>
<th>List</th>
<td><p><b>Local Birds</b>. Exclude birds unlikely to be found in your location.</p>
<p style="padding-left: 2em">With this option selected, an additional panel appears. The <b>Threshold</b> value represents the eBird checklist frequency.
So, a value of 0.03 means birds that are predicted to occur in less than 3% of eBird checlists in the area are excluded.
If you also check "Use Week", it excludes birds not expected at the time of year the recording was made - which is read from the files' timestamp.
</p>
<p><b>Nocturnal Birds</b>. Just report detections relating to these species' calls (song detections are
excluded).</p>
<p><b>All birds</b>. Include birdsong as well as calls, also includes species not known to call at night.
</p>
<p><b>Everything</b>. In addition to birds, show non-bird detections including segments that were considered noise.
</ul>
<p>If you wish to see the full list of the 320 species Chirpity <i>Nocmig</i> was trained on, just explore the list
offered when correcting a prediction.</p>
<p>If you wish to see the full list of species used for predictions each model was trained on, check the Help menu > <i>What species are detected?</i>.</p>
</td>
</tr>
<tr>
Expand All @@ -37,11 +52,48 @@
of the files and the location that is set.
</td>
</tr>
<tr>
<td colspan="2" class="text-center text-bg-light"><h5>System</h5></td>
</tr>
<tr>
<th>Language</th>
<td>Choose the language for the "common name" labels, where translations exist. The translated labels have been obtained from <a href="https://ebird.org/home" target="_blank">eBird</a>.
</td>
</tr>
<tr>
<th>Backend</th>
<td>Choose how predictions are calculated on you computer.
<ul>
<li>
<b>CPU</b> will use the CPU for predictions, this is generally faster if you have an integrated
graphics card.
</li>
<li>
<b>GPU</b> will use the graphics card. If you have a dedicated graphics card, this will typically be
significantly faster.
</li>
</ul>
</td>
</tr>

<tr>
<th>Threads</th>
<td>This setting allows you to select the number of threads to use for predictions.
When using the CPU backend, the number of threads is automatically set to match the number of CPU cores available on the computer.
By default, the GPU backend uses one thread. Adjusting this setting, along with the batch size, can optimize the speed of predictions.
</td>
</tr>
<tr>
<th>Batch Size</th>
<td>This setting determines the number of three-second audio chunks to process each time a prediction is requested.
Larger values lead to faster processing, especially on long files.
</td>
</tr>
<tr>
<td colspan="2" class="text-center text-bg-light"><h5>Location</h5></td>
</tr>
<tr>
<th>Location</th>
<th>Default Location</th>
<td>
<figure class="figure float-end m-3">
<img src="Help/amend file.png"
Expand All @@ -53,70 +105,48 @@
</td>
</tr>
<tr>
<td colspan="2" class="text-center text-bg-light"><h5>Spectrogram</h5></td>
<td colspan="2" class="text-center text-bg-light"><h5>Audio Export Preferences</h5></td>
</tr>
<tr>
<td><b>Colourmap</b></td>
<td>Choose the colour map for the spectrogram display</td>
<td><b>Format and Bitrate</b></td>
<td>Choose from lossless or lossy formats for the audio clips you export from Chirpity.</td>
</tr>
<tr>
<th>Timeline</th>
<td>The spectrogram timeline has two modes:
<th>Decorators</th>
<td>These two options can be used to:
<ul>
<li>
<b>Timecode</b> displays the time elapsed from the beginning of the file.
<b>Pad:</b> Add two seconds to the beginning and end of the audio selected for export
</li>
<li>
<b>Time of Day</b> relies on file timestamps to determine the actual time of detected calls.
If a file's timestamp does not reflect the recording time, the readonig will be inaccurate.
To edit the start time of the current file, set the desired date and time using the filename context menu.
<b>Fade:</b> Apply a fade effect to the beginning and end of your exported files.
</li>
</ul>
</td>
</tr>
<tr>
<td colspan="2" class="text-center text-bg-light"><h5>System</h5></td>
<td colspan="2" class="text-center text-bg-light"><h5>Spectrogram Preferences</h5></td>
</tr>
<tr>
<th>Backend</th>
<td>Choose how predictions are calculated on you computer.
<td><b>Colourmap</b></td>
<td>Choose the colour map for the spectrogram display</td>
</tr>
<tr>
<th>Timeline</th>
<td>The spectrogram timeline has two modes:
<ul>
<li>
<b>CPU</b> will use the CPU for predictions, this is generally faster if you have an integrated
graphics card.
<b>Timecode</b> displays the time elapsed from the beginning of the file.
</li>
<li>
<b>GPU</b> will use the graphics card. If you have a dedicated graphics card, this will typically be
significantly faster.
<b>Time of Day</b> relies on file timestamps to determine the actual time of detected calls.
If a file's timestamp does not reflect the recording time, the readonig will be inaccurate.
To edit the start time of the current file, set the desired date and time using the filename context menu.
</li>
</ul>
</td>
</tr>
<th>Model</th>
<td>Choose the model to use for prediction.
<ul>
<li>
<b>Chirpity</b> will use the native Chirpity model for predictions.
</li>
<li>
<b>BirdNET</b> will use the <a href="https://github.com/kahst/BirdNET-Analyzer" target="_blank">BirdNET</a> model developed by Stefan Kahl et al. from the K. Lisa Yang Center for Conservation Bioacoustics
</li>
</ul>
</td>
</tr>
<tr>
<th>Threads</th>
<td>This setting allows you to select the number of threads to use for predictions.
When using the CPU backend, the number of threads is automatically set to match the number of CPU cores available on the computer.
By default, the GPU backend uses one thread. Adjusting this setting, along with the batch size, can optimize the speed of predictions.
</td>
</tr>
<tr>
<th>Batch Size</th>
<td>This setting determines the number of three-second audio chunks to process each time a prediction is requested.
Larger values lead to faster processing, especially on long files.
</td>
</tr>

<tr>
<td colspan="2" class="text-center text-bg-warning"><h5>Experimental Features</h5></td>
</tr>
Expand Down
Loading

0 comments on commit 6567968

Please sign in to comment.