{
  "id": 239938,
  "title": "Compute power would be a major factor here",
  "url": "/competitions/siim-covid19-detection/discussion/239938",
  "author_name": "",
  "post_date": "2021-05-18T05:14:29.924312700Z",
  "votes": 3,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Looking at the size of the images, the 9 hrs GPU time for inference, I have a feeling this will be one of those competitions where high compute power would definitely give an edge. One could always work on a smaller image size but that would quite probably prevent one from getting to the top LB ranks.</p>\n<p>I still have to look more but these are my initial thoughts.</p>",
  "messages": [
    {
      "id": "1312515",
      "postDate": "05/18/2021 05:14:29",
      "content": "<p>Looking at the size of the images, the 9 hrs GPU time for inference, I have a feeling this will be one of those competitions where high compute power would definitely give an edge. One could always work on a smaller image size but that would quite probably prevent one from getting to the top LB ranks.</p>\n<p>I still have to look more but these are my initial thoughts.</p>",
      "rawMarkdown": "Looking at the size of the images, the 9 hrs GPU time for inference, I have a feeling this will be one of those competitions where high compute power would definitely give an edge. One could always work on a smaller image size but that would quite probably prevent one from getting to the top LB ranks.\n\nI still have to look more but these are my initial thoughts.",
      "votes": null
    },
    {
      "id": "1312565",
      "postDate": "05/18/2021 05:51:24",
      "content": "<p>6300 training samples only, DICOM is always large</p>",
      "rawMarkdown": "6300 training samples only, DICOM is always large",
      "votes": null
    },
    {
      "id": "1312587",
      "postDate": "05/18/2021 06:08:27",
      "content": "<p>Yes that is true. Let's see :)</p>",
      "rawMarkdown": "Yes that is true. Let's see :)",
      "votes": null
    },
    {
      "id": "1312698",
      "postDate": "05/18/2021 07:39:47",
      "content": "<p>There were already <a href=\"https://www.kaggle.com/c/rsna-str-pulmonary-embolism-detection\" target=\"_blank\">much bigger</a> kernel competition dataset</p>\n<p>And as said by <a href=\"https://www.kaggle.com/philippsinger\" target=\"_blank\">@philippsinger</a> Dicom contains info that majority here won't use.</p>",
      "rawMarkdown": "There were already [much bigger](https://www.kaggle.com/c/rsna-str-pulmonary-embolism-detection) kernel competition dataset\n\nAnd as said by @philippsinger Dicom contains info that majority here won't use.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1312565,
      "author_name": "philippsinger",
      "author_url": "",
      "post_date": "05/18/2021 05:51:24",
      "content": "<p>6300 training samples only, DICOM is always large</p>",
      "votes": null,
      "replies": [
        {
          "id": 1312587,
          "author_name": "shreyansh2626",
          "author_url": "",
          "post_date": "05/18/2021 06:08:27",
          "content": "<p>Yes that is true. Let's see :)</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1312698,
      "author_name": "serigne",
      "author_url": "",
      "post_date": "05/18/2021 07:39:47",
      "content": "<p>There were already <a href=\"https://www.kaggle.com/c/rsna-str-pulmonary-embolism-detection\" target=\"_blank\">much bigger</a> kernel competition dataset</p>\n<p>And as said by <a href=\"https://www.kaggle.com/philippsinger\" target=\"_blank\">@philippsinger</a> Dicom contains info that majority here won't use.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1312515": "Looking at the size of the images, the 9 hrs GPU time for inference, I have a feeling this will be one of those competitions where high compute power would definitely give an edge. One could always work on a smaller image size but that would quite probably prevent one from getting to the top LB ranks.\n\nI still have to look more but these are my initial thoughts.",
    "1312565": "6300 training samples only, DICOM is always large",
    "1312587": "Yes that is true. Let's see :)",
    "1312698": "There were already [much bigger](https://www.kaggle.com/c/rsna-str-pulmonary-embolism-detection) kernel competition dataset\n\nAnd as said by @philippsinger Dicom contains info that majority here won't use."
  },
  "source": "meta"
}