{
  "id": 206475,
  "title": "DCT2 transformed image as a CNN input for classification.",
  "url": "/competitions/cassava-leaf-disease-classification/discussion/206475",
  "author_name": "kildongGo",
  "post_date": "2020-12-24T20:05:33.261000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi, All.</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2737964%2F7e8572f5d9ba9529e3eae65bc8cc683e%2Fimgpsh_mobile_save-3.jpg?generation=1608840252938105&amp;alt=media\" alt=\"dct2\"></p>\n<p>-When the image is converted to DCT2, most of the geometric information is gathered near the top left.</p>\n<p>-For example, after converting a 1024x1024 ant image to DCT2, only the 256x256 ROI in the upper left corner is left, and the rest of the 1024x array filled with zeros can be reversed to restore almost like the original ant image.</p>\n<p>-The reason is that almost all geometric information is contained in the 256x256 DCT array.</p>\n<p>-Then, I ask you a question here. If you input the 256x256 DCT image in the upper left corner to CNN and train it, will the ant image be classified? Can't it?</p>\n<p>I wonder.</p>\n<p>Thanks.<br>\n<a href=\"https://www.kaggle.com/bemoregt\" target=\"_blank\">@bemoregt</a>.</p>",
  "messages": [
    {
      "id": 1125556,
      "postDate": "2020-12-24T20:05:33.260Z",
      "content": "<p>Hi, All.</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2737964%2F7e8572f5d9ba9529e3eae65bc8cc683e%2Fimgpsh_mobile_save-3.jpg?generation=1608840252938105&amp;alt=media\" alt=\"dct2\"></p>\n<p>-When the image is converted to DCT2, most of the geometric information is gathered near the top left.</p>\n<p>-For example, after converting a 1024x1024 ant image to DCT2, only the 256x256 ROI in the upper left corner is left, and the rest of the 1024x array filled with zeros can be reversed to restore almost like the original ant image.</p>\n<p>-The reason is that almost all geometric information is contained in the 256x256 DCT array.</p>\n<p>-Then, I ask you a question here. If you input the 256x256 DCT image in the upper left corner to CNN and train it, will the ant image be classified? Can't it?</p>\n<p>I wonder.</p>\n<p>Thanks.<br>\n<a href=\"https://www.kaggle.com/bemoregt\" target=\"_blank\">@bemoregt</a>.</p>",
      "rawMarkdown": "Hi, All.\n\n![dct2](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2737964%2F7e8572f5d9ba9529e3eae65bc8cc683e%2Fimgpsh_mobile_save-3.jpg?generation=1608840252938105&alt=media)\n\n-When the image is converted to DCT2, most of the geometric information is gathered near the top left.\n\n-For example, after converting a 1024x1024 ant image to DCT2, only the 256x256 ROI in the upper left corner is left, and the rest of the 1024x array filled with zeros can be reversed to restore almost like the original ant image.\n\n-The reason is that almost all geometric information is contained in the 256x256 DCT array.\n\n-Then, I ask you a question here. If you input the 256x256 DCT image in the upper left corner to CNN and train it, will the ant image be classified? Can't it?\n\nI wonder.\n\nThanks.\n@bemoregt."
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1125556": "Hi, All.\n\n![dct2](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2737964%2F7e8572f5d9ba9529e3eae65bc8cc683e%2Fimgpsh_mobile_save-3.jpg?generation=1608840252938105&alt=media)\n\n-When the image is converted to DCT2, most of the geometric information is gathered near the top left.\n\n-For example, after converting a 1024x1024 ant image to DCT2, only the 256x256 ROI in the upper left corner is left, and the rest of the 1024x array filled with zeros can be reversed to restore almost like the original ant image.\n\n-The reason is that almost all geometric information is contained in the 256x256 DCT array.\n\n-Then, I ask you a question here. If you input the 256x256 DCT image in the upper left corner to CNN and train it, will the ant image be classified? Can't it?\n\nI wonder.\n\nThanks.\n@bemoregt."
  }
}