{
  "id": 269772,
  "title": "Still cannot understand how I can load sequential images to the model.",
  "url": "/competitions/rsna-miccai-brain-tumor-radiogenomic-classification/discussion/269772",
  "author_name": "",
  "post_date": "2021-09-02T03:46:04.487168200Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>First of all I would like to thank <a href=\"https://www.kaggle.com/returnofsputnik\" target=\"_blank\">@returnofsputnik</a> for helping me resolve the previous issue I had. He also gave some pretty good ideas. But there are certain things that I still am not able to understand. If I am padding the images wouldn't that affect the performance? for instance suppose I have 28 images in FLAIR and say only 4 images in T1w. Then I'll have to pad those four images so that I can match the dimensions. Can you guys please give me ideas on how you were able to load the data into your model for training? Thank you!</p>",
  "messages": [
    {
      "id": "1499919",
      "postDate": "09/02/2021 03:46:04",
      "content": "<p>First of all I would like to thank <a href=\"https://www.kaggle.com/returnofsputnik\" target=\"_blank\">@returnofsputnik</a> for helping me resolve the previous issue I had. He also gave some pretty good ideas. But there are certain things that I still am not able to understand. If I am padding the images wouldn't that affect the performance? for instance suppose I have 28 images in FLAIR and say only 4 images in T1w. Then I'll have to pad those four images so that I can match the dimensions. Can you guys please give me ideas on how you were able to load the data into your model for training? Thank you!</p>",
      "rawMarkdown": "First of all I would like to thank @returnofsputnik for helping me resolve the previous issue I had. He also gave some pretty good ideas. But there are certain things that I still am not able to understand. If I am padding the images wouldn't that affect the performance? for instance suppose I have 28 images in FLAIR and say only 4 images in T1w. Then I'll have to pad those four images so that I can match the dimensions. Can you guys please give me ideas on how you were able to load the data into your model for training? Thank you!",
      "votes": null
    },
    {
      "id": "1500328",
      "postDate": "09/02/2021 09:37:46",
      "content": "<p>From <a href=\"https://www.kaggle.com/experienceinai/aivo-vii-nopea\" target=\"_blank\">https://www.kaggle.com/experienceinai/aivo-vii-nopea</a> you can find \"initial notebook\" code to dig in to folders and picking the images from there … matching the dimensions etc. Resulting e.g. to a form: </p>\n<p>Opetustiedon 0 muoto:  (640, 128, 128, 1)<br>\nOpetuksen vastetiedon muoto:  (640, 2)</p>\n<p>(Opetus = training)</p>\n<p>…meaning you have 640 cases of images size 128x128, easy to feed in to your own CNN. By investigating the code you can select which folders you want to use (only one or all four etc…). Hope this helps you in the beginning!</p>",
      "rawMarkdown": "From https://www.kaggle.com/experienceinai/aivo-vii-nopea you can find \"initial notebook\" code to dig in to folders and picking the images from there ... matching the dimensions etc. Resulting e.g. to a form: \n\nOpetustiedon 0 muoto:  (640, 128, 128, 1)\nOpetuksen vastetiedon muoto:  (640, 2)\n\n(Opetus = training)\n\n...meaning you have 640 cases of images size 128x128, easy to feed in to your own CNN. By investigating the code you can select which folders you want to use (only one or all four etc...). Hope this helps you in the beginning!",
      "votes": null
    },
    {
      "id": "1500343",
      "postDate": "09/02/2021 09:52:02",
      "content": "<p>Thank you, I'll take a look at that notebook :D</p>",
      "rawMarkdown": "Thank you, I'll take a look at that notebook :D",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1500328,
      "author_name": "experienceinai",
      "author_url": "",
      "post_date": "09/02/2021 09:37:46",
      "content": "<p>From <a href=\"https://www.kaggle.com/experienceinai/aivo-vii-nopea\" target=\"_blank\">https://www.kaggle.com/experienceinai/aivo-vii-nopea</a> you can find \"initial notebook\" code to dig in to folders and picking the images from there … matching the dimensions etc. Resulting e.g. to a form: </p>\n<p>Opetustiedon 0 muoto:  (640, 128, 128, 1)<br>\nOpetuksen vastetiedon muoto:  (640, 2)</p>\n<p>(Opetus = training)</p>\n<p>…meaning you have 640 cases of images size 128x128, easy to feed in to your own CNN. By investigating the code you can select which folders you want to use (only one or all four etc…). Hope this helps you in the beginning!</p>",
      "votes": null,
      "replies": [
        {
          "id": 1500343,
          "author_name": "eddwait",
          "author_url": "",
          "post_date": "09/02/2021 09:52:02",
          "content": "<p>Thank you, I'll take a look at that notebook :D</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1499919": "First of all I would like to thank @returnofsputnik for helping me resolve the previous issue I had. He also gave some pretty good ideas. But there are certain things that I still am not able to understand. If I am padding the images wouldn't that affect the performance? for instance suppose I have 28 images in FLAIR and say only 4 images in T1w. Then I'll have to pad those four images so that I can match the dimensions. Can you guys please give me ideas on how you were able to load the data into your model for training? Thank you!",
    "1500328": "From https://www.kaggle.com/experienceinai/aivo-vii-nopea you can find \"initial notebook\" code to dig in to folders and picking the images from there ... matching the dimensions etc. Resulting e.g. to a form: \n\nOpetustiedon 0 muoto:  (640, 128, 128, 1)\nOpetuksen vastetiedon muoto:  (640, 2)\n\n(Opetus = training)\n\n...meaning you have 640 cases of images size 128x128, easy to feed in to your own CNN. By investigating the code you can select which folders you want to use (only one or all four etc...). Hope this helps you in the beginning!",
    "1500343": "Thank you, I'll take a look at that notebook :D"
  },
  "source": "meta"
}