{
  "id": 178446,
  "title": "I tried data augmentation, but the results still the same! 😨",
  "url": "/competitions/digit-recognizer/discussion/178446",
  "author_name": "PyBear",
  "post_date": "2020-08-29T23:12:10.039000",
  "votes": 1,
  "comment_count": 3,
  "views": null,
  "content": "<p>Can some who has some expertise with data augmentation to help me?</p>",
  "messages": [
    {
      "id": 990958,
      "postDate": "2020-08-30T02:00:23.857Z",
      "content": "<p>Hello !</p>\n<p>What type of data augmentation did you do ? It pains me to admit it but first time I tried some I actually went way too far with it, I rotate certain images almost 180 degres like an idiot, without realizing the 9 will end up looking like 6 :D.</p>\n<p>I would strongly suggest this amazing notebook by the amazing 4xGrandmaster Chris Deotte : <a href=\"https://www.kaggle.com/cdeotte/25-million-images-0-99757-mnist\" target=\"_blank\">link</a>. It's actually quite easy to understand, and he used data augmentation at the beginning (you have to click some of the 'code' buttons to see more parts of it).</p>\n<p>Good training !</p>",
      "rawMarkdown": "Hello !\n\nWhat type of data augmentation did you do ? It pains me to admit it but first time I tried some I actually went way too far with it, I rotate certain images almost 180 degres like an idiot, without realizing the 9 will end up looking like 6 :D.\n\nI would strongly suggest this amazing notebook by the amazing 4xGrandmaster Chris Deotte : [link](https://www.kaggle.com/cdeotte/25-million-images-0-99757-mnist). It's actually quite easy to understand, and he used data augmentation at the beginning (you have to click some of the 'code' buttons to see more parts of it).\n\nGood training !",
      "votes": 1,
      "replies": [
        {
          "id": 991260,
          "postDate": "2020-08-30T09:16:39.423Z",
          "content": "<p>Thanks Vincent.</p>",
          "rawMarkdown": "Thanks Vincent.",
          "votes": 1
        }
      ]
    },
    {
      "id": 990861,
      "postDate": "2020-08-29T23:12:10.040Z",
      "content": "<p>Can some who has some expertise with data augmentation to help me?</p>",
      "rawMarkdown": "Can some who has some expertise with data augmentation to help me?",
      "votes": 1
    },
    {
      "id": 996476,
      "postDate": "2020-09-03T10:21:51.963Z",
      "content": "<p>Hi check out <a href=\"https://www.kaggle.com/alankritamishra/learncnn?rvi=1\" target=\"_blank\">https://www.kaggle.com/alankritamishra/learncnn?rvi=1</a> it might help</p>",
      "rawMarkdown": "Hi check out https://www.kaggle.com/alankritamishra/learncnn?rvi=1 it might help"
    }
  ],
  "comments": [
    {
      "id": 990958,
      "author_name": "Vincent Dion",
      "author_url": "",
      "post_date": "2020-08-30T02:00:23.857000",
      "content": "<p>Hello !</p>\n<p>What type of data augmentation did you do ? It pains me to admit it but first time I tried some I actually went way too far with it, I rotate certain images almost 180 degres like an idiot, without realizing the 9 will end up looking like 6 :D.</p>\n<p>I would strongly suggest this amazing notebook by the amazing 4xGrandmaster Chris Deotte : <a href=\"https://www.kaggle.com/cdeotte/25-million-images-0-99757-mnist\" target=\"_blank\">link</a>. It's actually quite easy to understand, and he used data augmentation at the beginning (you have to click some of the 'code' buttons to see more parts of it).</p>\n<p>Good training !</p>",
      "votes": 1,
      "replies": [
        {
          "id": 991260,
          "author_name": "PyBear",
          "author_url": "",
          "post_date": "2020-08-30T09:16:39.423000",
          "content": "<p>Thanks Vincent.</p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 996476,
      "author_name": "Alankrita Mishra",
      "author_url": "",
      "post_date": "2020-09-03T10:21:51.963000",
      "content": "<p>Hi check out <a href=\"https://www.kaggle.com/alankritamishra/learncnn?rvi=1\" target=\"_blank\">https://www.kaggle.com/alankritamishra/learncnn?rvi=1</a> it might help</p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "990958": "Hello !\n\nWhat type of data augmentation did you do ? It pains me to admit it but first time I tried some I actually went way too far with it, I rotate certain images almost 180 degres like an idiot, without realizing the 9 will end up looking like 6 :D.\n\nI would strongly suggest this amazing notebook by the amazing 4xGrandmaster Chris Deotte : [link](https://www.kaggle.com/cdeotte/25-million-images-0-99757-mnist). It's actually quite easy to understand, and he used data augmentation at the beginning (you have to click some of the 'code' buttons to see more parts of it).\n\nGood training !",
    "990861": "Can some who has some expertise with data augmentation to help me?",
    "996476": "Hi check out https://www.kaggle.com/alankritamishra/learncnn?rvi=1 it might help"
  }
}