{
  "id": 369813,
  "title": "Discussion on the training pipeline for multiple images",
  "url": "/competitions/rsna-breast-cancer-detection/discussion/369813",
  "author_name": "Jacky S",
  "post_date": "2022-12-01T14:01:02.949000",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>After the EDA i have done myself, I found there are quite a lot of patients that have more than 4 images taken.  <br>\nSo far i have no idea how to build the pipeline for fitting varing numbers of images.</p>\n<p>Any idea?</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F5836519%2Ffa560a6808540dc283e1c7259146a558%2FScreenshot%202022-12-01%20at%209.58.13%20PM.png?generation=1669903113015609&amp;alt=media\" alt=\"No of Images for patients\"></p>\n<p>Link to the notebook: <a href=\"https://www.kaggle.com/code/jackysywk/newbie-exploratory-data-analysis\" target=\"_blank\">Exploratory Data Analysis</a></p>",
  "messages": [
    {
      "id": 2052264,
      "postDate": "2022-12-02T03:39:33.457Z",
      "content": "<p><a href=\"https://paperswithcode.com/task/multiple-instance-learning\" target=\"_blank\">Multiple Instance Learning</a> is always worth trying out in such contexts 🙂 Here, we would need to group data in 'bags' based on how labels are assigned (we have a label per breast, hence that is how we would need to group the images as well).</p>\n<p>Would be fun to try it out!</p>",
      "rawMarkdown": "[Multiple Instance Learning](https://paperswithcode.com/task/multiple-instance-learning) is always worth trying out in such contexts 🙂 Here, we would need to group data in 'bags' based on how labels are assigned (we have a label per breast, hence that is how we would need to group the images as well).\n\nWould be fun to try it out!",
      "votes": 1
    },
    {
      "id": 2051577,
      "postDate": "2022-12-01T14:01:02.950Z",
      "content": "<p>After the EDA i have done myself, I found there are quite a lot of patients that have more than 4 images taken.  <br>\nSo far i have no idea how to build the pipeline for fitting varing numbers of images.</p>\n<p>Any idea?</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F5836519%2Ffa560a6808540dc283e1c7259146a558%2FScreenshot%202022-12-01%20at%209.58.13%20PM.png?generation=1669903113015609&amp;alt=media\" alt=\"No of Images for patients\"></p>\n<p>Link to the notebook: <a href=\"https://www.kaggle.com/code/jackysywk/newbie-exploratory-data-analysis\" target=\"_blank\">Exploratory Data Analysis</a></p>",
      "rawMarkdown": "After the EDA i have done myself, I found there are quite a lot of patients that have more than 4 images taken.  \nSo far i have no idea how to build the pipeline for fitting varing numbers of images.\n\nAny idea?\n\n![No of Images for patients](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F5836519%2Ffa560a6808540dc283e1c7259146a558%2FScreenshot%202022-12-01%20at%209.58.13%20PM.png?generation=1669903113015609&alt=media)\n\n\nLink to the notebook: [Exploratory Data Analysis](https://www.kaggle.com/code/jackysywk/newbie-exploratory-data-analysis)",
      "votes": 1
    }
  ],
  "comments": [
    {
      "id": 2052264,
      "author_name": "Radek Osmulski",
      "author_url": "",
      "post_date": "2022-12-02T03:39:33.457000",
      "content": "<p><a href=\"https://paperswithcode.com/task/multiple-instance-learning\" target=\"_blank\">Multiple Instance Learning</a> is always worth trying out in such contexts 🙂 Here, we would need to group data in 'bags' based on how labels are assigned (we have a label per breast, hence that is how we would need to group the images as well).</p>\n<p>Would be fun to try it out!</p>",
      "votes": 1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2052264": "[Multiple Instance Learning](https://paperswithcode.com/task/multiple-instance-learning) is always worth trying out in such contexts 🙂 Here, we would need to group data in 'bags' based on how labels are assigned (we have a label per breast, hence that is how we would need to group the images as well).\n\nWould be fun to try it out!",
    "2051577": "After the EDA i have done myself, I found there are quite a lot of patients that have more than 4 images taken.  \nSo far i have no idea how to build the pipeline for fitting varing numbers of images.\n\nAny idea?\n\n![No of Images for patients](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F5836519%2Ffa560a6808540dc283e1c7259146a558%2FScreenshot%202022-12-01%20at%209.58.13%20PM.png?generation=1669903113015609&alt=media)\n\n\nLink to the notebook: [Exploratory Data Analysis](https://www.kaggle.com/code/jackysywk/newbie-exploratory-data-analysis)"
  }
}