{
  "id": 402579,
  "title": "Time periods (5s) in sample_submission.csv file",
  "url": "/competitions/birdclef-2023/discussion/402579",
  "author_name": "kreative24hk",
  "post_date": "2023-04-18T23:00:06.846000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I have a question about the sample_submission.csv rows.</p>\n<p>In the example, it divides the test file into 5sec windows and predicts each bird's likelihood in that time frame. I am curious whether it should be 5sec or not. Does it matter if we can set sample_submission.csv rows like using 10sec windows, not 5sec windows?</p>\n<p>I am not sure I understand the evaluation policy correctly but, It doesn't matter how many rows we have in our sample_submission.csv. The thing we should keep in mind is just the number of columns same as 264 classes, right? If yes, during the evaluation, will the likelihood of each bird be finally calculated automatically by averaging column-wise?</p>",
  "messages": [
    {
      "id": 2226435,
      "postDate": "2023-04-18T23:00:06.847Z",
      "content": "<p>I have a question about the sample_submission.csv rows.</p>\n<p>In the example, it divides the test file into 5sec windows and predicts each bird's likelihood in that time frame. I am curious whether it should be 5sec or not. Does it matter if we can set sample_submission.csv rows like using 10sec windows, not 5sec windows?</p>\n<p>I am not sure I understand the evaluation policy correctly but, It doesn't matter how many rows we have in our sample_submission.csv. The thing we should keep in mind is just the number of columns same as 264 classes, right? If yes, during the evaluation, will the likelihood of each bird be finally calculated automatically by averaging column-wise?</p>",
      "rawMarkdown": "I have a question about the sample_submission.csv rows.\n\nIn the example, it divides the test file into 5sec windows and predicts each bird's likelihood in that time frame. I am curious whether it should be 5sec or not. Does it matter if we can set sample_submission.csv rows like using 10sec windows, not 5sec windows?\n\nI am not sure I understand the evaluation policy correctly but, It doesn't matter how many rows we have in our sample_submission.csv. The thing we should keep in mind is just the number of columns same as 264 classes, right? If yes, during the evaluation, will the likelihood of each bird be finally calculated automatically by averaging column-wise?\n\n"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2226435": "I have a question about the sample_submission.csv rows.\n\nIn the example, it divides the test file into 5sec windows and predicts each bird's likelihood in that time frame. I am curious whether it should be 5sec or not. Does it matter if we can set sample_submission.csv rows like using 10sec windows, not 5sec windows?\n\nI am not sure I understand the evaluation policy correctly but, It doesn't matter how many rows we have in our sample_submission.csv. The thing we should keep in mind is just the number of columns same as 264 classes, right? If yes, during the evaluation, will the likelihood of each bird be finally calculated automatically by averaging column-wise?\n\n"
  }
}