{
  "id": 168647,
  "title": "Does multiple augmentations on dataset helps in increasing LB score ?",
  "url": "/competitions/siim-isic-melanoma-classification/discussion/168647",
  "author_name": "",
  "post_date": "2020-07-21T11:22:30.184927300Z",
  "votes": null,
  "comment_count": 4,
  "views": 0,
  "content": "",
  "messages": [
    {
      "id": "938144",
      "postDate": "07/21/2020 11:22:30",
      "content": "",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "938328",
      "postDate": "07/21/2020 13:08:31",
      "content": "<p>Yes, multiple augmentations increase the model's ability to generalize, and avoids over-fitting to the training dataset, thereby increasing the final Private LB score. (This is the same reason for using Dropout Regularization in Neural Networks.)</p>",
      "rawMarkdown": "Yes, multiple augmentations increase the model's ability to generalize, and avoids over-fitting to the training dataset, thereby increasing the final Private LB score. (This is the same reason for using Dropout Regularization in Neural Networks.)",
      "votes": null
    },
    {
      "id": "938414",
      "postDate": "07/21/2020 14:06:30",
      "content": "<p>what could be some of the useful augmentations in this competition, if you can tell me?</p>",
      "rawMarkdown": "what could be some of the useful augmentations in this competition, if you can tell me?",
      "votes": null
    },
    {
      "id": "938425",
      "postDate": "07/21/2020 14:16:36",
      "content": "<ul>\n<li>Rotations</li>\n<li>Flips</li>\n<li>Zoom</li>\n<li>Shear</li>\n<li>Saturation</li>\n<li>Contrast</li>\n<li>Brightness</li>\n</ul>\n\n<p>Are ones that I've found useful in this competition</p>",
      "rawMarkdown": "Rotations\n- Flips\n- Zoom\n- Shear\n- Saturation\n- Contrast\n- Brightness\n\nAre ones that I've found useful in this competition",
      "votes": null
    },
    {
      "id": "939203",
      "postDate": "07/22/2020 05:04:11",
      "content": "<p>Along with augmentations suggested by <a href=\"/santiviquez\">@santiviquez</a> you can also add Hair augmentation and shift augmentation which works well.</p>",
      "rawMarkdown": "Along with augmentations suggested by @santiviquez you can also add Hair augmentation and shift augmentation which works well.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 938328,
      "author_name": "sirishks",
      "author_url": "",
      "post_date": "07/21/2020 13:08:31",
      "content": "<p>Yes, multiple augmentations increase the model's ability to generalize, and avoids over-fitting to the training dataset, thereby increasing the final Private LB score. (This is the same reason for using Dropout Regularization in Neural Networks.)</p>",
      "votes": null,
      "replies": [
        {
          "id": 938414,
          "author_name": "prashantarorat",
          "author_url": "",
          "post_date": "07/21/2020 14:06:30",
          "content": "<p>what could be some of the useful augmentations in this competition, if you can tell me?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 938425,
          "author_name": "santiviquez",
          "author_url": "",
          "post_date": "07/21/2020 14:16:36",
          "content": "<ul>\n<li>Rotations</li>\n<li>Flips</li>\n<li>Zoom</li>\n<li>Shear</li>\n<li>Saturation</li>\n<li>Contrast</li>\n<li>Brightness</li>\n</ul>\n\n<p>Are ones that I've found useful in this competition</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 939203,
          "author_name": "prateek0x",
          "author_url": "",
          "post_date": "07/22/2020 05:04:11",
          "content": "<p>Along with augmentations suggested by <a href=\"/santiviquez\">@santiviquez</a> you can also add Hair augmentation and shift augmentation which works well.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "938144": "",
    "938328": "Yes, multiple augmentations increase the model's ability to generalize, and avoids over-fitting to the training dataset, thereby increasing the final Private LB score. (This is the same reason for using Dropout Regularization in Neural Networks.)",
    "938414": "what could be some of the useful augmentations in this competition, if you can tell me?",
    "938425": "Rotations\n- Flips\n- Zoom\n- Shear\n- Saturation\n- Contrast\n- Brightness\n\nAre ones that I've found useful in this competition",
    "939203": "Along with augmentations suggested by @santiviquez you can also add Hair augmentation and shift augmentation which works well."
  },
  "source": "meta"
}