{
  "id": 169223,
  "title": "How to avoid bias?",
  "url": "/competitions/global-wheat-detection/discussion/169223",
  "author_name": "",
  "post_date": "2020-07-23T09:00:28.138600100Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Greetings of the day,\nWhat would be the solution to avoid high bias?\nThanks</p>",
  "messages": [
    {
      "id": "941512",
      "postDate": "07/23/2020 09:00:28",
      "content": "<p>Greetings of the day,\nWhat would be the solution to avoid high bias?\nThanks</p>",
      "rawMarkdown": "Greetings of the day,\nWhat would be the solution to avoid high bias?\nThanks",
      "votes": null
    },
    {
      "id": "942868",
      "postDate": "07/24/2020 03:37:00",
      "content": "<p>A basic solution is to apply many different kinds of augmentations that present the image as if it were taken under different conditions (height from ground by zooming/cropping, camera/lighting by adjusting brightness and color, etc). This will allow your model to generalize better to unseen images and prevent overfitting of the training images.</p>\n\n<p>There are certainly more advanced approaches though. Domain generalization and adaptation are terms used a lot in this context. I have gathered several resources of approaches to domain generalization <a href=\"https://www.kaggle.com/c/global-wheat-detection/discussion/161541\">here</a>.</p>",
      "rawMarkdown": "A basic solution is to apply many different kinds of augmentations that present the image as if it were taken under different conditions (height from ground by zooming/cropping, camera/lighting by adjusting brightness and color, etc). This will allow your model to generalize better to unseen images and prevent overfitting of the training images.\n\nThere are certainly more advanced approaches though. Domain generalization and adaptation are terms used a lot in this context. I have gathered several resources of approaches to domain generalization [here](https://www.kaggle.com/c/global-wheat-detection/discussion/161541).",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 942868,
      "author_name": "matthewmasters",
      "author_url": "",
      "post_date": "07/24/2020 03:37:00",
      "content": "<p>A basic solution is to apply many different kinds of augmentations that present the image as if it were taken under different conditions (height from ground by zooming/cropping, camera/lighting by adjusting brightness and color, etc). This will allow your model to generalize better to unseen images and prevent overfitting of the training images.</p>\n\n<p>There are certainly more advanced approaches though. Domain generalization and adaptation are terms used a lot in this context. I have gathered several resources of approaches to domain generalization <a href=\"https://www.kaggle.com/c/global-wheat-detection/discussion/161541\">here</a>.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "941512": "Greetings of the day,\nWhat would be the solution to avoid high bias?\nThanks",
    "942868": "A basic solution is to apply many different kinds of augmentations that present the image as if it were taken under different conditions (height from ground by zooming/cropping, camera/lighting by adjusting brightness and color, etc). This will allow your model to generalize better to unseen images and prevent overfitting of the training images.\n\nThere are certainly more advanced approaches though. Domain generalization and adaptation are terms used a lot in this context. I have gathered several resources of approaches to domain generalization [here](https://www.kaggle.com/c/global-wheat-detection/discussion/161541)."
  },
  "source": "meta"
}