{
  "id": 322456,
  "title": "Anyone can explain why training on smaller size then inference on bigger works?",
  "url": "/competitions/tensorflow-great-barrier-reef/discussion/322456",
  "author_name": "",
  "post_date": "2022-05-02T08:48:33.082142500Z",
  "votes": 6,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I am going through top solutions, and found a lot use large resolution, what's more, the resolution varies from training to inference, how that works? what's the mechanism?</p>",
  "messages": [
    {
      "id": "1774529",
      "postDate": "05/02/2022 08:48:33",
      "content": "<p>I am going through top solutions, and found a lot use large resolution, what's more, the resolution varies from training to inference, how that works? what's the mechanism?</p>",
      "rawMarkdown": "I am going through top solutions, and found a lot use large resolution, what's more, the resolution varies from training to inference, how that works? what's the mechanism?",
      "votes": null
    },
    {
      "id": "2022804",
      "postDate": "11/09/2022 09:49:19",
      "content": "<p>I think the test set has a lot of small objects. Increasing the resolution can help the model find those small objects, as they are all enlarged in the images.</p>",
      "rawMarkdown": "I think the test set has a lot of small objects. Increasing the resolution can help the model find those small objects, as they are all enlarged in the images.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2022804,
      "author_name": "yangly94",
      "author_url": "",
      "post_date": "11/09/2022 09:49:19",
      "content": "<p>I think the test set has a lot of small objects. Increasing the resolution can help the model find those small objects, as they are all enlarged in the images.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1774529": "I am going through top solutions, and found a lot use large resolution, what's more, the resolution varies from training to inference, how that works? what's the mechanism?",
    "2022804": "I think the test set has a lot of small objects. Increasing the resolution can help the model find those small objects, as they are all enlarged in the images."
  },
  "source": "meta"
}