{
  "id": 309855,
  "title": "A Question about MAP@5",
  "url": "/competitions/happy-whale-and-dolphin/discussion/309855",
  "author_name": "",
  "post_date": "2022-02-26T02:17:38.212403Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>This is my first time to participate in a kaggle competition, and I'm a little confused about the evaluation. I find the content from <a href=\"url\" target=\"_blank\">https://www.kaggle.com/pestipeti/explanation-of-map5-scoring-metric</a> very helpful. However, I still don't understand <strong>how to generate mutiple predictions per image. Does that mean I have to train multiple models?</strong><br>\nMaybe this sounds a little trivial for some of you, but it'll be great if you can help. <br>\nThank you.🙏</p>",
  "messages": [
    {
      "id": "1704945",
      "postDate": "02/26/2022 02:17:38",
      "content": "<p>This is my first time to participate in a kaggle competition, and I'm a little confused about the evaluation. I find the content from <a href=\"url\" target=\"_blank\">https://www.kaggle.com/pestipeti/explanation-of-map5-scoring-metric</a> very helpful. However, I still don't understand <strong>how to generate mutiple predictions per image. Does that mean I have to train multiple models?</strong><br>\nMaybe this sounds a little trivial for some of you, but it'll be great if you can help. <br>\nThank you.🙏</p>",
      "rawMarkdown": "This is my first time to participate in a kaggle competition, and I'm a little confused about the evaluation. I find the content from [https://www.kaggle.com/pestipeti/explanation-of-map5-scoring-metric](url) very helpful. However, I still don't understand **how to generate mutiple predictions per image. Does that mean I have to train multiple models?**\nMaybe this sounds a little trivial for some of you, but it'll be great if you can help. \nThank you.🙏",
      "votes": null
    },
    {
      "id": "1704970",
      "postDate": "02/26/2022 02:58:04",
      "content": "<p>From one model, you can take the top five predictions. For example, if you use embeddings, you predict the top five closest embeddings for the test image through the use of an algorithm like K-Nearest Neighbors. </p>",
      "rawMarkdown": "From one model, you can take the top five predictions. For example, if you use embeddings, you predict the top five closest embeddings for the test image through the use of an algorithm like K-Nearest Neighbors.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1704970,
      "author_name": "vexxingbanana",
      "author_url": "",
      "post_date": "02/26/2022 02:58:04",
      "content": "<p>From one model, you can take the top five predictions. For example, if you use embeddings, you predict the top five closest embeddings for the test image through the use of an algorithm like K-Nearest Neighbors. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1704945": "This is my first time to participate in a kaggle competition, and I'm a little confused about the evaluation. I find the content from [https://www.kaggle.com/pestipeti/explanation-of-map5-scoring-metric](url) very helpful. However, I still don't understand **how to generate mutiple predictions per image. Does that mean I have to train multiple models?**\nMaybe this sounds a little trivial for some of you, but it'll be great if you can help. \nThank you.🙏",
    "1704970": "From one model, you can take the top five predictions. For example, if you use embeddings, you predict the top five closest embeddings for the test image through the use of an algorithm like K-Nearest Neighbors."
  },
  "source": "meta"
}