{ "id": 242638, "title": "Question about the train_soundscapes", "url": "/competitions/birdclef-2021/discussion/242638", "author_name": "", "post_date": "2021-05-30T00:53:31.794169Z", "votes": 7, "comment_count": 14, "views": 0, "content": "

Hiii there, sorry if I miss this info earlier but are train_soundscapes recordings taken from the test set or are they completely separate/independent from the test set. Thank you!

", "messages": [ { "id": "1328151", "postDate": "05/30/2021 00:53:31", "content": "

Hiii there, sorry if I miss this info earlier but are train_soundscapes recordings taken from the test set or are they completely separate/independent from the test set. Thank you!

", "rawMarkdown": "Hiii there, sorry if I miss this info earlier but are train_soundscapes recordings taken from the test set or are they completely separate/independent from the test set. Thank you!", "votes": null }, { "id": "1329092", "postDate": "05/30/2021 20:49:15", "content": "

Completely different from test set.

", "rawMarkdown": "Completely different from test set.", "votes": null }, { "id": "1329484", "postDate": "05/31/2021 07:16:45", "content": "

How do you know it is completely different? As many people see CV on train soundscapes is matching with public LB.

", "rawMarkdown": "How do you know it is completely different? As many people see CV on train soundscapes is matching with public LB.", "votes": null }, { "id": "1329494", "postDate": "05/31/2021 07:25:54", "content": "
\n

As many people see CV on train soundscapes is matching with public LB.

\n
\n

If true, then why the question in the post?

", "rawMarkdown": "> As many people see CV on train soundscapes is matching with public LB. \n\nIf true, then why the question in the post?", "votes": null }, { "id": "1329552", "postDate": "05/31/2021 07:58:50", "content": "

There are 40 birds in the train soundscapes, 40 of 397 from what to predict. And also, in the same format as the test set. So, it will/should correlate is my guess but is only an indication of just that kind of bird calls of the specific bird, no one knows what weights and bird calls to expect in the private set.

", "rawMarkdown": "There are 40 birds in the train soundscapes, 40 of 397 from what to predict. And also, in the same format as the test set. So, it will/should correlate is my guess but is only an indication of just that kind of bird calls of the specific bird, no one knows what weights and bird calls to expect in the private set.", "votes": null }, { "id": "1329580", "postDate": "05/31/2021 08:19:49", "content": "

Are you sure it is 40?

", "rawMarkdown": "Are you sure it is 40?", "votes": null }, { "id": "1329588", "postDate": "05/31/2021 08:26:11", "content": "

No not now ;) just did a if birds in classes count code. Thanks, will check again.

", "rawMarkdown": "No not now ;) just did a if birds in classes count code. Thanks, will check again.", "votes": null }, { "id": "1329623", "postDate": "05/31/2021 08:58:08", "content": "

Did a check again and that’s what is says, and I didn't count the nocall just the birds, so one can count in two ways :)

", "rawMarkdown": "Did a check again and that’s what is says, and I didn't count the nocall just the birds, so one can count in two ways :)", "votes": null }, { "id": "1330660", "postDate": "06/01/2021 03:37:59", "content": "

I count 48 species in train soundscape

", "rawMarkdown": "I count 48 species in train soundscape", "votes": null }, { "id": "1330962", "postDate": "06/01/2021 07:36:10", "content": "

Yes correct yukiya, did a quick search with this first but didn't got them all within row wise without splitting first

\n
hits = 0\nwclass = []\nfor ii in range(len(CLASSES)):\n    if CLASSES[ii] in train_soundscape:\n        hits = hits+1\n        wclass.append(CLASSES[ii])\n\n        print(hits,CLASSES[ii])\n
\n

This did the job though

\n
wclass = []\nfor ii in range(len(train_soundscape)):\n    for splitbird in train_soundscape[ii].split():\n        if splitbird in CLASSES:\n            wclass.append(splitbird)\n\nprint(pd.DataFrame(wclass)[0].unique())\nprint(len(pd.DataFrame(wclass)[0].unique()))\n
\n

Glad for the extra validation birds ;) but more validations are needed I guess.

", "rawMarkdown": "Yes correct yukiya, did a quick search with this first but didn't got them all within row wise without splitting first\n\n```\nhits = 0\nwclass = []\nfor ii in range(len(CLASSES)):\n if CLASSES[ii] in train_soundscape:\n hits = hits+1\n wclass.append(CLASSES[ii])\n \n print(hits,CLASSES[ii])\n```\n\nThis did the job though\n\n```\nwclass = []\nfor ii in range(len(train_soundscape)):\n for splitbird in train_soundscape[ii].split():\n if splitbird in CLASSES:\n wclass.append(splitbird)\n\nprint(pd.DataFrame(wclass)[0].unique())\nprint(len(pd.DataFrame(wclass)[0].unique()))\n```\nGlad for the extra validation birds ;) but more validations are needed I guess.", "votes": null }, { "id": "1331091", "postDate": "06/01/2021 09:10:54", "content": "

48 is not 40 ;)

", "rawMarkdown": "48 is not 40 ;)", "votes": null }, { "id": "1331122", "postDate": "06/01/2021 09:40:55", "content": "

But 40 in range(48) ;) Yes correct, showed above in the code where the differences came from. Thanks for the help and first notice :) 👍
\nBtw, will be a great reading your write-up/solution, what a score, feels miles away.

", "rawMarkdown": "But 40 in range(48) ;) Yes correct, showed above in the code where the differences came from. Thanks for the help and first notice :) 👍 \nBtw, will be a great reading your write-up/solution, what a score, feels miles away.", "votes": null }, { "id": "1331201", "postDate": "06/01/2021 10:29:27", "content": "

Thanks, but I'll wait to see what private score is before claiming anything is great ;)

", "rawMarkdown": "Thanks, but I'll wait to see what private score is before claiming anything is great ;)", "votes": null }, { "id": "1332064", "postDate": "06/01/2021 21:54:07", "content": "

@cpmpml u expect any shake up from experience. although cv correlate for many inspite of that there were s hake ups before .

", "rawMarkdown": "cpmpml u expect any shake up from experience. although cv correlate for many inspite of that there were s hake ups before .", "votes": null }, { "id": "1332122", "postDate": "06/01/2021 23:38:16", "content": "

It depend son how public/private test is split. If public only has COR and SSW sites (same as train soundscape), then huge shakeup is possible.

\n

If not then shakeup could be small. Saying so because my CV LB correlation is very good.

", "rawMarkdown": "It depend son how public/private test is split. If public only has COR and SSW sites (same as train soundscape), then huge shakeup is possible. \n\nIf not then shakeup could be small. Saying so because my CV LB correlation is very good.", "votes": null } ], "comments": [ { "id": 1329092, "author_name": "zaber666", "author_url": "", "post_date": "05/30/2021 20:49:15", "content": "

Completely different from test set.

", "votes": null, "replies": [ { "id": 1329484, "author_name": "superchenhao", "author_url": "", "post_date": "05/31/2021 07:16:45", "content": "

How do you know it is completely different? As many people see CV on train soundscapes is matching with public LB.

", "votes": null, "replies": [] }, { "id": 1329494, "author_name": "cpmpml", "author_url": "", "post_date": "05/31/2021 07:25:54", "content": "
\n

As many people see CV on train soundscapes is matching with public LB.

\n
\n

If true, then why the question in the post?

", "votes": null, "replies": [] } ] }, { "id": 1329552, "author_name": "kirderf", "author_url": "", "post_date": "05/31/2021 07:58:50", "content": "

There are 40 birds in the train soundscapes, 40 of 397 from what to predict. And also, in the same format as the test set. So, it will/should correlate is my guess but is only an indication of just that kind of bird calls of the specific bird, no one knows what weights and bird calls to expect in the private set.

", "votes": null, "replies": [ { "id": 1329580, "author_name": "cpmpml", "author_url": "", "post_date": "05/31/2021 08:19:49", "content": "

Are you sure it is 40?

", "votes": null, "replies": [] }, { "id": 1329588, "author_name": "kirderf", "author_url": "", "post_date": "05/31/2021 08:26:11", "content": "

No not now ;) just did a if birds in classes count code. Thanks, will check again.

", "votes": null, "replies": [] }, { "id": 1329623, "author_name": "kirderf", "author_url": "", "post_date": "05/31/2021 08:58:08", "content": "

Did a check again and that’s what is says, and I didn't count the nocall just the birds, so one can count in two ways :)

", "votes": null, "replies": [] }, { "id": 1330660, "author_name": "nyleve", "author_url": "", "post_date": "06/01/2021 03:37:59", "content": "

I count 48 species in train soundscape

", "votes": null, "replies": [] }, { "id": 1330962, "author_name": "kirderf", "author_url": "", "post_date": "06/01/2021 07:36:10", "content": "

Yes correct yukiya, did a quick search with this first but didn't got them all within row wise without splitting first

\n
hits = 0\nwclass = []\nfor ii in range(len(CLASSES)):\n    if CLASSES[ii] in train_soundscape:\n        hits = hits+1\n        wclass.append(CLASSES[ii])\n\n        print(hits,CLASSES[ii])\n
\n

This did the job though

\n
wclass = []\nfor ii in range(len(train_soundscape)):\n    for splitbird in train_soundscape[ii].split():\n        if splitbird in CLASSES:\n            wclass.append(splitbird)\n\nprint(pd.DataFrame(wclass)[0].unique())\nprint(len(pd.DataFrame(wclass)[0].unique()))\n
\n

Glad for the extra validation birds ;) but more validations are needed I guess.

", "votes": null, "replies": [] }, { "id": 1331091, "author_name": "cpmpml", "author_url": "", "post_date": "06/01/2021 09:10:54", "content": "

48 is not 40 ;)

", "votes": null, "replies": [] }, { "id": 1331122, "author_name": "kirderf", "author_url": "", "post_date": "06/01/2021 09:40:55", "content": "

But 40 in range(48) ;) Yes correct, showed above in the code where the differences came from. Thanks for the help and first notice :) 👍
\nBtw, will be a great reading your write-up/solution, what a score, feels miles away.

", "votes": null, "replies": [] }, { "id": 1331201, "author_name": "cpmpml", "author_url": "", "post_date": "06/01/2021 10:29:27", "content": "

Thanks, but I'll wait to see what private score is before claiming anything is great ;)

", "votes": null, "replies": [] }, { "id": 1332064, "author_name": "jaideepvalani", "author_url": "", "post_date": "06/01/2021 21:54:07", "content": "

@cpmpml u expect any shake up from experience. although cv correlate for many inspite of that there were s hake ups before .

", "votes": null, "replies": [] }, { "id": 1332122, "author_name": "cpmpml", "author_url": "", "post_date": "06/01/2021 23:38:16", "content": "

It depend son how public/private test is split. If public only has COR and SSW sites (same as train soundscape), then huge shakeup is possible.

\n

If not then shakeup could be small. Saying so because my CV LB correlation is very good.

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1328151": "Hiii there, sorry if I miss this info earlier but are train_soundscapes recordings taken from the test set or are they completely separate/independent from the test set. Thank you!", "1329092": "Completely different from test set.", "1329484": "How do you know it is completely different? As many people see CV on train soundscapes is matching with public LB.", "1329494": "> As many people see CV on train soundscapes is matching with public LB. \n\nIf true, then why the question in the post?", "1329552": "There are 40 birds in the train soundscapes, 40 of 397 from what to predict. And also, in the same format as the test set. So, it will/should correlate is my guess but is only an indication of just that kind of bird calls of the specific bird, no one knows what weights and bird calls to expect in the private set.", "1329580": "Are you sure it is 40?", "1329588": "No not now ;) just did a if birds in classes count code. Thanks, will check again.", "1329623": "Did a check again and that’s what is says, and I didn't count the nocall just the birds, so one can count in two ways :)", "1330660": "I count 48 species in train soundscape", "1330962": "Yes correct yukiya, did a quick search with this first but didn't got them all within row wise without splitting first\n\n```\nhits = 0\nwclass = []\nfor ii in range(len(CLASSES)):\n if CLASSES[ii] in train_soundscape:\n hits = hits+1\n wclass.append(CLASSES[ii])\n \n print(hits,CLASSES[ii])\n```\n\nThis did the job though\n\n```\nwclass = []\nfor ii in range(len(train_soundscape)):\n for splitbird in train_soundscape[ii].split():\n if splitbird in CLASSES:\n wclass.append(splitbird)\n\nprint(pd.DataFrame(wclass)[0].unique())\nprint(len(pd.DataFrame(wclass)[0].unique()))\n```\nGlad for the extra validation birds ;) but more validations are needed I guess.", "1331091": "48 is not 40 ;)", "1331122": "But 40 in range(48) ;) Yes correct, showed above in the code where the differences came from. Thanks for the help and first notice :) 👍 \nBtw, will be a great reading your write-up/solution, what a score, feels miles away.", "1331201": "Thanks, but I'll wait to see what private score is before claiming anything is great ;)", "1332064": "cpmpml u expect any shake up from experience. although cv correlate for many inspite of that there were s hake ups before .", "1332122": "It depend son how public/private test is split. If public only has COR and SSW sites (same as train soundscape), then huge shakeup is possible. \n\nIf not then shakeup could be small. Saying so because my CV LB correlation is very good." }, "source": "meta" }