{
  "id": 177254,
  "title": "example test audio summary not aligned with metadata start end",
  "url": "/competitions/birdsong-recognition/discussion/177254",
  "author_name": "",
  "post_date": "2020-08-25T09:39:27.448261600Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Although there are only 9 ebird codes in the example test audio clips, in testing have found that the label birds in the summary are not always (perhaps rarely) aligned with the metadata start/end times.  <br>\ne.g.  summary for BLKFR-10-CPL_20190611_093000_20 &amp; _25 have mouchi but in the metadata  mouchi start ends are 17.1395792  17.77564895   &amp;    19.58600132    20.02635731<br>\nat 25 - 30 in metadata  ebird codes are gockin and westan. </p>\n<p>So if using this for any testing or comparisons with train, it may be better to create clips out of metadata than use summary.  But not sure how reliable that is either. </p>",
  "messages": [
    {
      "id": "984792",
      "postDate": "08/25/2020 09:39:27",
      "content": "<p>Although there are only 9 ebird codes in the example test audio clips, in testing have found that the label birds in the summary are not always (perhaps rarely) aligned with the metadata start/end times.  <br>\ne.g.  summary for BLKFR-10-CPL_20190611_093000_20 &amp; _25 have mouchi but in the metadata  mouchi start ends are 17.1395792  17.77564895   &amp;    19.58600132    20.02635731<br>\nat 25 - 30 in metadata  ebird codes are gockin and westan. </p>\n<p>So if using this for any testing or comparisons with train, it may be better to create clips out of metadata than use summary.  But not sure how reliable that is either. </p>",
      "rawMarkdown": "Although there are only 9 ebird codes in the example test audio clips, in testing have found that the label birds in the summary are not always (perhaps rarely) aligned with the metadata start/end times.  \ne.g.  summary for BLKFR-10-CPL_20190611_093000_20 & _25 have mouchi but in the metadata  mouchi start ends are 17.1395792  17.77564895   &    19.58600132\t20.02635731\nat 25 - 30 in metadata  ebird codes are gockin and westan. \n\nSo if using this for any testing or comparisons with train, it may be better to create clips out of metadata than use summary.  But not sure how reliable that is either.",
      "votes": null
    },
    {
      "id": "985011",
      "postDate": "08/25/2020 12:41:33",
      "content": "<p>For R users to see time clips in Audacity:</p>\n<p>library(data.table)<br>\nlibrary(tidyverse)<br>\nlibrary(seewave)<br>\ntmet&lt;- fread(\"C:\\bird\\example_test_audio_metadata.csv\")<br>\ntmet&lt;-filter(tmet,device==\"BLKFR-10-CPL\")</p>\n<h1>or</h1>\n<h1>tmet&lt;-filter(tmet,device==\"ORANGE-7-CAP\")</h1>\n<p>label &lt;- substr(tmet$ebird_code,1,1)<br>\nt1 &lt;- tmet$time_start<br>\nt2 &lt;- tmet$time_end<br>\ndf &lt;- data.frame(label, t1, t2)<br>\nwrite.audacity(df, filename=\"clip.txt\")</p>\n<p>Then in Audacity:<br>\nFile-import-labels-clip.txt</p>\n<p>Effect-Amplify for low level signals</p>",
      "rawMarkdown": "For R users to see time clips in Audacity:\n\nlibrary(data.table)\nlibrary(tidyverse)\nlibrary(seewave)\ntmet<- fread(\"C:\\\\bird\\\\example_test_audio_metadata.csv\")\ntmet<-filter(tmet,device==\"BLKFR-10-CPL\")\n# or\n# tmet<-filter(tmet,device==\"ORANGE-7-CAP\")\nlabel <- substr(tmet$ebird_code,1,1)\nt1 <- tmet$time_start\nt2 <- tmet$time_end\ndf <- data.frame(label, t1, t2)\nwrite.audacity(df, filename=\"clip.txt\")\n\nThen in Audacity:\nFile-import-labels-clip.txt\n\nEffect-Amplify for low level signals",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 985011,
      "author_name": "tomkkk",
      "author_url": "",
      "post_date": "08/25/2020 12:41:33",
      "content": "<p>For R users to see time clips in Audacity:</p>\n<p>library(data.table)<br>\nlibrary(tidyverse)<br>\nlibrary(seewave)<br>\ntmet&lt;- fread(\"C:\\bird\\example_test_audio_metadata.csv\")<br>\ntmet&lt;-filter(tmet,device==\"BLKFR-10-CPL\")</p>\n<h1>or</h1>\n<h1>tmet&lt;-filter(tmet,device==\"ORANGE-7-CAP\")</h1>\n<p>label &lt;- substr(tmet$ebird_code,1,1)<br>\nt1 &lt;- tmet$time_start<br>\nt2 &lt;- tmet$time_end<br>\ndf &lt;- data.frame(label, t1, t2)<br>\nwrite.audacity(df, filename=\"clip.txt\")</p>\n<p>Then in Audacity:<br>\nFile-import-labels-clip.txt</p>\n<p>Effect-Amplify for low level signals</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "984792": "Although there are only 9 ebird codes in the example test audio clips, in testing have found that the label birds in the summary are not always (perhaps rarely) aligned with the metadata start/end times.  \ne.g.  summary for BLKFR-10-CPL_20190611_093000_20 & _25 have mouchi but in the metadata  mouchi start ends are 17.1395792  17.77564895   &    19.58600132\t20.02635731\nat 25 - 30 in metadata  ebird codes are gockin and westan. \n\nSo if using this for any testing or comparisons with train, it may be better to create clips out of metadata than use summary.  But not sure how reliable that is either.",
    "985011": "For R users to see time clips in Audacity:\n\nlibrary(data.table)\nlibrary(tidyverse)\nlibrary(seewave)\ntmet<- fread(\"C:\\\\bird\\\\example_test_audio_metadata.csv\")\ntmet<-filter(tmet,device==\"BLKFR-10-CPL\")\n# or\n# tmet<-filter(tmet,device==\"ORANGE-7-CAP\")\nlabel <- substr(tmet$ebird_code,1,1)\nt1 <- tmet$time_start\nt2 <- tmet$time_end\ndf <- data.frame(label, t1, t2)\nwrite.audacity(df, filename=\"clip.txt\")\n\nThen in Audacity:\nFile-import-labels-clip.txt\n\nEffect-Amplify for low level signals"
  },
  "source": "meta"
}