{ "id": 170749, "title": "[caution][for beginners] librosa.load()", "url": "/competitions/birdsong-recognition/discussion/170749", "author_name": "", "post_date": "2020-07-28T22:43:28.933682100Z", "votes": 8, "comment_count": 4, "views": 0, "content": "

Hello everyone.

\n\n

I'm going to share something that confused me when I was using librosa.load().
\nAs has been pointed out in several notebooks, you need to be careful when using librosa.load().

\n\n

See the following example.

\n\n

python\npath_audio = '. /train_audio/aldfly/XC134874.mp3'\nsignal, sr = librosa.load(path_audio)\nprint(f'sampling rate: {sr}')
\n>> sampling rate: 22050
\n

\n\n

If you don't give an argument of sr, librosa.load() will load the audio file with the default value of sr=22050.\nThis means that the sampling rate is not accurate.
\nI'll check the train.csv for the sampling rate.

\n\n

python\nsr = df_train[df_train.filename=='XC134874.mp3']['sampling_rate'][0]\nprint(sr)
\n>> '48000 (Hz)'
\n

\n\n

The exact sampling rate of the audio file seems to be 48000 Hz. librosa.load() can easily work around this problem by giving sr=None as an argument.

\n\n

python\nsignal, sr = librosa.load(path_audio, sr=None)\nprint(f'sampling rate: {sr}')
\n>> sampling rate: 48000\n

\n\n

> ref: http://man.hubwiz.com/docset/LibROSA.docset/Contents/Resources/Documents/generated/librosa.core.load.html?highlight=load#librosa.core.load

\n\n

That's it!

\n\n

arigato !!

\n\n
\n\n

Update

\n\n

@hidehisaarai1213 commented. (Thank you @hidehisaarai1213 !! )

\n\n
\n

If the original sr and given sr does not match, it performs resampling.

\n
", "messages": [ { "id": "949768", "postDate": "07/28/2020 22:43:28", "content": "

Hello everyone.

\n\n

I'm going to share something that confused me when I was using librosa.load().
\nAs has been pointed out in several notebooks, you need to be careful when using librosa.load().

\n\n

See the following example.

\n\n

python\npath_audio = '. /train_audio/aldfly/XC134874.mp3'\nsignal, sr = librosa.load(path_audio)\nprint(f'sampling rate: {sr}')
\n>> sampling rate: 22050
\n

\n\n

If you don't give an argument of sr, librosa.load() will load the audio file with the default value of sr=22050.\nThis means that the sampling rate is not accurate.
\nI'll check the train.csv for the sampling rate.

\n\n

python\nsr = df_train[df_train.filename=='XC134874.mp3']['sampling_rate'][0]\nprint(sr)
\n>> '48000 (Hz)'
\n

\n\n

The exact sampling rate of the audio file seems to be 48000 Hz. librosa.load() can easily work around this problem by giving sr=None as an argument.

\n\n

python\nsignal, sr = librosa.load(path_audio, sr=None)\nprint(f'sampling rate: {sr}')
\n>> sampling rate: 48000\n

\n\n

> ref: http://man.hubwiz.com/docset/LibROSA.docset/Contents/Resources/Documents/generated/librosa.core.load.html?highlight=load#librosa.core.load

\n\n

That's it!

\n\n

arigato !!

\n\n
\n\n

Update

\n\n

@hidehisaarai1213 commented. (Thank you @hidehisaarai1213 !! )

\n\n
\n

If the original sr and given sr does not match, it performs resampling.

\n
", "rawMarkdown": "### Hello everyone. \nI'm going to share something that **confused me** when I was using `librosa.load()`. \nAs has been pointed out in several notebooks, you need to be careful when using `librosa.load()`. \n\nSee the following example.\n\n```python\npath_audio = '. /train_audio/aldfly/XC134874.mp3'\nsignal, sr = librosa.load(path_audio)\nprint(f'sampling rate: {sr}') \n>> sampling rate: 22050 \n```\n\nIf you don't give an argument of sr, librosa.load() will load the audio file with the default value of sr=22050.\nThis means that the sampling rate is not accurate. \nI'll check the `train.csv` for the sampling rate.\n\n```python\nsr = df_train[df_train.filename=='XC134874.mp3']['sampling_rate'][0]\nprint(sr) \n>> '48000 (Hz)' \n```\n\nThe exact sampling rate of the audio file seems to be 48000 Hz. librosa.load() can easily work around this problem by giving `sr=None` as an argument.\n\n```python\nsignal, sr = librosa.load(path_audio, sr=None)\nprint(f'sampling rate: {sr}') \n>> sampling rate: 48000\n```\n\n\n> ref: http://man.hubwiz.com/docset/LibROSA.docset/Contents/Resources/Documents/generated/librosa.core.load.html?highlight=load#librosa.core.load\n\n### That's it! \n### arigato !!\n\n--- \n\n### Update \n@hidehisaarai1213 commented. (Thank you @hidehisaarai1213 !! )\n> If the original sr and given sr does not match, it performs resampling.", "votes": null }, { "id": "949784", "postDate": "07/28/2020 23:42:39", "content": "

If the original sr and given sr does not match, it performs resampling.

", "rawMarkdown": "If the original sr and given sr does not match, it performs resampling.", "votes": null }, { "id": "949857", "postDate": "07/29/2020 02:46:19", "content": "

@hidehisaarai1213

\n\n

I guess I should add that to the post. \nI'll add that later.\nThank you !!

", "rawMarkdown": "hidehisaarai1213 \n\nI guess I should add that to the post. \nI'll add that later.\nThank you !!", "votes": null }, { "id": "950003", "postDate": "07/29/2020 05:49:40", "content": "

Or you specify None as SR to preserve the original one

", "rawMarkdown": "Or you specify None as SR to preserve the original one", "votes": null }, { "id": "950841", "postDate": "07/29/2020 17:00:05", "content": "

Thank you

", "rawMarkdown": "Thank you", "votes": null } ], "comments": [ { "id": 949784, "author_name": "hidehisaarai1213", "author_url": "", "post_date": "07/28/2020 23:42:39", "content": "

If the original sr and given sr does not match, it performs resampling.

", "votes": null, "replies": [ { "id": 949857, "author_name": "fkubota", "author_url": "", "post_date": "07/29/2020 02:46:19", "content": "

@hidehisaarai1213

\n\n

I guess I should add that to the post. \nI'll add that later.\nThank you !!

", "votes": null, "replies": [] }, { "id": 950003, "author_name": "snovik1975", "author_url": "", "post_date": "07/29/2020 05:49:40", "content": "

Or you specify None as SR to preserve the original one

", "votes": null, "replies": [] } ] }, { "id": 950841, "author_name": "priyanshujain218", "author_url": "", "post_date": "07/29/2020 17:00:05", "content": "

Thank you

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "949768": "### Hello everyone. \nI'm going to share something that **confused me** when I was using `librosa.load()`. \nAs has been pointed out in several notebooks, you need to be careful when using `librosa.load()`. \n\nSee the following example.\n\n```python\npath_audio = '. /train_audio/aldfly/XC134874.mp3'\nsignal, sr = librosa.load(path_audio)\nprint(f'sampling rate: {sr}') \n>> sampling rate: 22050 \n```\n\nIf you don't give an argument of sr, librosa.load() will load the audio file with the default value of sr=22050.\nThis means that the sampling rate is not accurate. \nI'll check the `train.csv` for the sampling rate.\n\n```python\nsr = df_train[df_train.filename=='XC134874.mp3']['sampling_rate'][0]\nprint(sr) \n>> '48000 (Hz)' \n```\n\nThe exact sampling rate of the audio file seems to be 48000 Hz. librosa.load() can easily work around this problem by giving `sr=None` as an argument.\n\n```python\nsignal, sr = librosa.load(path_audio, sr=None)\nprint(f'sampling rate: {sr}') \n>> sampling rate: 48000\n```\n\n\n> ref: http://man.hubwiz.com/docset/LibROSA.docset/Contents/Resources/Documents/generated/librosa.core.load.html?highlight=load#librosa.core.load\n\n### That's it! \n### arigato !!\n\n--- \n\n### Update \n@hidehisaarai1213 commented. (Thank you @hidehisaarai1213 !! )\n> If the original sr and given sr does not match, it performs resampling.", "949784": "If the original sr and given sr does not match, it performs resampling.", "949857": "hidehisaarai1213 \n\nI guess I should add that to the post. \nI'll add that later.\nThank you !!", "950003": "Or you specify None as SR to preserve the original one", "950841": "Thank you" }, "source": "meta" }