{
  "id": 205883,
  "title": "I've made DCT cut-out augmentation method.",
  "url": "/competitions/cassava-leaf-disease-classification/discussion/205883",
  "author_name": "kildongGo",
  "post_date": "2020-12-22T10:28:16.266000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi, All.</p>\n<p>I've made DCT cut-out augmentation method.</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2737964%2Fa35570f4ef7b59e214ab7df41fe6dcc6%2FScrShot%2020.png?generation=1608632863562089&amp;alt=media\" alt=\"I've made DCT cut-out augmentation method.\"></p>\n<p>The DCT means Discrete Cosine Transform.</p>\n<p>Animated GIF Image(before &amp; after) is here:<br>\n<a href=\"https://www.facebook.com/100008007287779/videos/2716983841911832/\" target=\"_blank\">https://www.facebook.com/100008007287779/videos/2716983841911832/</a></p>\n<p>It's procedure is here:</p>\n<ol>\n<li>2D DCT transform of grayscale image.</li>\n<li>Cut out small box in DCT domain array.</li>\n<li>Inverse 2D DCT transform of cut-outed DCT domain array. </li>\n</ol>\n<p>Is this a good augmentation for image classification's generalization performance?</p>\n<p>Thanks.<br>\nBest,<br>\n<a href=\"https://www.kaggle.com/bemoregt\" target=\"_blank\">@bemoregt</a>.</p>",
  "messages": [
    {
      "id": 1122280,
      "postDate": "2020-12-22T10:28:16.267Z",
      "content": "<p>Hi, All.</p>\n<p>I've made DCT cut-out augmentation method.</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2737964%2Fa35570f4ef7b59e214ab7df41fe6dcc6%2FScrShot%2020.png?generation=1608632863562089&amp;alt=media\" alt=\"I've made DCT cut-out augmentation method.\"></p>\n<p>The DCT means Discrete Cosine Transform.</p>\n<p>Animated GIF Image(before &amp; after) is here:<br>\n<a href=\"https://www.facebook.com/100008007287779/videos/2716983841911832/\" target=\"_blank\">https://www.facebook.com/100008007287779/videos/2716983841911832/</a></p>\n<p>It's procedure is here:</p>\n<ol>\n<li>2D DCT transform of grayscale image.</li>\n<li>Cut out small box in DCT domain array.</li>\n<li>Inverse 2D DCT transform of cut-outed DCT domain array. </li>\n</ol>\n<p>Is this a good augmentation for image classification's generalization performance?</p>\n<p>Thanks.<br>\nBest,<br>\n<a href=\"https://www.kaggle.com/bemoregt\" target=\"_blank\">@bemoregt</a>.</p>",
      "rawMarkdown": "Hi, All.\n\nI've made DCT cut-out augmentation method.\n\n![I've made DCT cut-out augmentation method.](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2737964%2Fa35570f4ef7b59e214ab7df41fe6dcc6%2FScrShot%2020.png?generation=1608632863562089&alt=media)\n\nThe DCT means Discrete Cosine Transform.\n\nAnimated GIF Image(before & after) is here:\nhttps://www.facebook.com/100008007287779/videos/2716983841911832/\n\n\n\nIt's procedure is here:\n\n1. 2D DCT transform of grayscale image.\n2. Cut out small box in DCT domain array.\n3. Inverse 2D DCT transform of cut-outed DCT domain array. \n\nIs this a good augmentation for image classification's generalization performance?\n\nThanks.\nBest,\n@bemoregt."
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1122280": "Hi, All.\n\nI've made DCT cut-out augmentation method.\n\n![I've made DCT cut-out augmentation method.](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2737964%2Fa35570f4ef7b59e214ab7df41fe6dcc6%2FScrShot%2020.png?generation=1608632863562089&alt=media)\n\nThe DCT means Discrete Cosine Transform.\n\nAnimated GIF Image(before & after) is here:\nhttps://www.facebook.com/100008007287779/videos/2716983841911832/\n\n\n\nIt's procedure is here:\n\n1. 2D DCT transform of grayscale image.\n2. Cut out small box in DCT domain array.\n3. Inverse 2D DCT transform of cut-outed DCT domain array. \n\nIs this a good augmentation for image classification's generalization performance?\n\nThanks.\nBest,\n@bemoregt."
  }
}