Moved frequency weighting image to be included in package

CHANGELOG.md

@@ -12,4 +12,9 @@ Removed device assignment that was causing errors on distributed training setups
 ## 0.1.2 (2024-05-16)
 
 #### Bug Fix
-Error in tensor shapes. Was processing as the incorrect [batch, channels, stem, samples] instead of the correct [batch, stem, channels, samples] in some places.
+Error in tensor shapes. Was processing as the incorrect [batch, channels, stem, samples] instead of the correct [batch, stem, channels, samples] in some places.
+
+## 0.1.3 (2024-05-16)
+
+#### Include Image
+Moved frequency weighting image to be included in the package.

README.md

@@ -65,7 +65,7 @@ The goal of this metric is to account for several factors not present in current
 ##### Frequency Weighting
 To measure the frequencies of a signal closer to that of human hearing, the following frequency weighting is applied. This helps the model effectively pay less attention to errors at frequencies that humans are not sensitive to (e.g. 50 Hz) and give more weight to those that we are acutely tuned to (e.g. 3kHz).
 
-![Frequency Weighting](images/frequency_weighting.png)
+![Frequency Weighting](torch_log_wmse_audio_quality/images/frequency_weighting.png)
 
 This metric has been constructed with high fidelity audio in mind (sample rates ≥ 44.1kHz). It theoretically could work for lower sample rates, like 16kHz, but the metric performs an internal resampling to 44.1kHz for consistency across any input sample rates.
 

setup.cfg

@@ -1,6 +1,6 @@
 [metadata]
 name = torch-log-wmse-audio-quality
-version = 0.1.2
+version = 0.1.3
 author = Christopher Landschoot
 author_email = [email protected]
 license = Apache License 2.0
@@ -25,7 +25,7 @@ install_requires =
     numpy>=1.23.5
 
 [options.package_data]
-torch_log_wmse_audio_quality = filter_ir.pkl
+torch_log_wmse_audio_quality = filter_ir.pkl, images/frequency_weighting.png
 
 [options.packages.find]
 where = .