Feature/evaluation #146
Feature/evaluation #146
Conversation
# Conflicts: # pytest/modules/UltraStar/test_ultrastar_writer.py # requirements.txt # src/Settings.py # src/UltraSinger.py # src/modules/Audio/separation.py # src/modules/Pitcher/pitcher.py # src/modules/Speech_Recognition/TranscribedData.py # src/modules/Speech_Recognition/Whisper.py # src/modules/Ultrastar/ultrastar_parser.py # src/modules/Ultrastar/ultrastar_txt.py # src/modules/Ultrastar/ultrastar_writer.py
pytest/modules/UltraSinger.py
Outdated
|
|
||
| class PitcherTest(unittest.TestCase): | ||
| # @pytest.mark.skip(reason="Skipping this FUNCTION level test, can be used for manual tests") | ||
| def test_get_pitch_with_crepe_file(self): |
There was a problem hiding this comment.
This test can be removed. PitcherTest class is in test_pitcher.py
| import pytest | ||
| from src.modules.plot import plot | ||
|
|
||
|
|
||
| class PitcherTest(unittest.TestCase): | ||
| @pytest.mark.skip(reason="Skipping this FUNCTION level test, can be used for manual tests") | ||
| # @pytest.mark.skip(reason="Skipping this FUNCTION level test, can be used for manual tests") |
| @@ -1,6 +1,7 @@ | |||
| MIT License | |||
|
|
|||
| Copyright (c) 2023 Vadim Rangnau | |||
| Copyright (c) 2020 Max Morrison (torchcrepe code adapted for crepe output filtering abd thresholding) | |||
There was a problem hiding this comment.
maybe fork crepe an make a separate package for it? Or PR the changes to the crepe project?
pytest/modules/UltraSinger.py
Outdated
| from src.modules.plot import plot | ||
|
|
||
|
|
||
| class PitcherTest(unittest.TestCase): |
There was a problem hiding this comment.
This is also in test_pitcher.py
| @@ -1,6 +1,7 @@ | |||
| MIT License | |||
|
|
|||
| Copyright (c) 2023 Vadim Rangnau | |||
| Copyright (c) 2020 Max Morrison (torchcrepe code adapted for crepe output filtering abd thresholding) | |||
There was a problem hiding this comment.
Why not fork crepe, move the changes there and release an package?
There was a problem hiding this comment.
I wasn't sure how to give proper credit here. I would say it's up to you how to handle this and if you want to accept the code in question.
The code here https://github.com/rakuri255/UltraSinger/pull/146/files#diff-7bda13ea2689179c7952b68174b8b8ea2cc2250f9b9699c5cf55939fb6c4ac7d is copied and adapted from here https://github.com/maxrmorrison/torchcrepe/blob/master/torchcrepe/loudness.py
| @@ -881,6 +1024,19 @@ def denoise_vocal_audio(input_path: str, output_path: str) -> None: | |||
| """Denoise vocal audio""" | |||
| ffmpeg_reduce_noise(input_path, output_path) | |||
|
|
|||
| # Fixme: Merge issue | |||
src/modules/console_colors.py
Outdated
| @@ -8,6 +8,11 @@ def blue_highlighted(text: str) -> str: | |||
| return f"{Bcolors.blue}{text}{Bcolors.endc}" | |||
|
|
|||
|
|
|||
| def green_highlighted(text: str) -> str: | |||
| """Returns a blue highlighted text""" | |||
| FILE_ENCODING, | ||
| ) | ||
|
|
||
| CHARACTERS_TO_REMOVE = ["\ufeff"] |
There was a problem hiding this comment.
Is this "BOM"? Than we should name it "BOM"
| ultrastar_class = UltrastarTxtValue() | ||
| count = 0 | ||
|
|
||
| # Strips the newline character | ||
| for line in txt: | ||
| filtered_line = line | ||
| for character_to_remove in CHARACTERS_TO_REMOVE: |
There was a problem hiding this comment.
Why remove in each line? Isnt this just an encoding issue?
|
|
||
| @dataclass_json | ||
| @dataclass | ||
| class TranscriptionResult: |
There was a problem hiding this comment.
maybe just Transcription with data and detected_language?
# Conflicts: # src/UltraSinger.py # src/modules/Speech_Recognition/TranscribedData.py # src/modules/plot.py
| start_time = ( | ||
| beat_to_second(int(ultrastar_class.startBeat[pos]), real_bpm) + gap | ||
| ) | ||
| gap = int(float(ultrastar_class.gap.replace(",", ".")) / 1000) |
There was a problem hiding this comment.
This breaks the score calculation
| real_bpm = ultrastar_bpm_to_real_bpm( | ||
| float(ultrastar_class.bpm.replace(",", ".")) | ||
| ) | ||
| gap = int(float(ultrastar_class.gap.replace(",", ".")) / 1000) |
There was a problem hiding this comment.
This breaks the score calculation
There was a problem hiding this comment.
good catch.
another contributing factor to broken score calculation was the wrong usage of the constructor here:
76a4956#diff-f8287eea965b098ba5cef0e0b678bd57ea9d41c3985c3a311c940792de15c57fR987
Calling the constructor like this
new_transcribed_data.append(TranscribedData({"word": midi_segment.word, "start": midi_segment.start, "end": midi_segment.end, "is_hyphen": None, "confidence": 1}))just initializes the confidence field to have the map as it's value and all other fields will have their default value. So all start and end values are 0 from then on.
No description provided.