{
  "id": 162532,
  "title": "How to get training data?",
  "url": "/competitions/birdsong-recognition/discussion/162532",
  "author_name": "",
  "post_date": "2020-06-29T08:04:46.154794900Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>In my understanding, We have a lot of MP3 training data, they vary from one minute to two minutes, and the bird label is about MP3.However, in test set, we need to predict whether there is bird call in some 5s.Did I get it wrong?</p>",
  "messages": [
    {
      "id": "906349",
      "postDate": "06/29/2020 08:04:46",
      "content": "<p>In my understanding, We have a lot of MP3 training data, they vary from one minute to two minutes, and the bird label is about MP3.However, in test set, we need to predict whether there is bird call in some 5s.Did I get it wrong?</p>",
      "rawMarkdown": "In my understanding, We have a lot of MP3 training data, they vary from one minute to two minutes, and the bird label is about MP3.However, in test set, we need to predict whether there is bird call in some 5s.Did I get it wrong?",
      "votes": null
    },
    {
      "id": "906445",
      "postDate": "06/29/2020 09:48:05",
      "content": "<p>There are two types of test set audio, Set 1 and Set 2 in which you need to predict every 5 seconds, while in Set 3 you need to predict on the whole audio, there are a few guides in the notebook section on how to predict on the hidden test set, here is one of most simpler one so easily adaptable to your code: <a href=\"https://www.kaggle.com/cwthompson/birdsong-making-a-prediction\">Prediction Notebook</a> or this one <a href=\"https://www.kaggle.com/hidehisaarai1213/inference-pytorch-birdcall-resnet-baseline\">Notebook 2</a> in which you can also see how to create a basic model.</p>",
      "rawMarkdown": "There are two types of test set audio, Set 1 and Set 2 in which you need to predict every 5 seconds, while in Set 3 you need to predict on the whole audio, there are a few guides in the notebook section on how to predict on the hidden test set, here is one of most simpler one so easily adaptable to your code: [Prediction Notebook](https://www.kaggle.com/cwthompson/birdsong-making-a-prediction) or this one [Notebook 2](https://www.kaggle.com/hidehisaarai1213/inference-pytorch-birdcall-resnet-baseline) in which you can also see how to create a basic model.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 906445,
      "author_name": "pranavkasela",
      "author_url": "",
      "post_date": "06/29/2020 09:48:05",
      "content": "<p>There are two types of test set audio, Set 1 and Set 2 in which you need to predict every 5 seconds, while in Set 3 you need to predict on the whole audio, there are a few guides in the notebook section on how to predict on the hidden test set, here is one of most simpler one so easily adaptable to your code: <a href=\"https://www.kaggle.com/cwthompson/birdsong-making-a-prediction\">Prediction Notebook</a> or this one <a href=\"https://www.kaggle.com/hidehisaarai1213/inference-pytorch-birdcall-resnet-baseline\">Notebook 2</a> in which you can also see how to create a basic model.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "906349": "In my understanding, We have a lot of MP3 training data, they vary from one minute to two minutes, and the bird label is about MP3.However, in test set, we need to predict whether there is bird call in some 5s.Did I get it wrong?",
    "906445": "There are two types of test set audio, Set 1 and Set 2 in which you need to predict every 5 seconds, while in Set 3 you need to predict on the whole audio, there are a few guides in the notebook section on how to predict on the hidden test set, here is one of most simpler one so easily adaptable to your code: [Prediction Notebook](https://www.kaggle.com/cwthompson/birdsong-making-a-prediction) or this one [Notebook 2](https://www.kaggle.com/hidehisaarai1213/inference-pytorch-birdcall-resnet-baseline) in which you can also see how to create a basic model."
  },
  "source": "meta"
}