{
  "id": 214603,
  "title": "How to visualize 4 channels in a single image? ",
  "url": "/competitions/hpa-single-cell-image-classification/discussion/214603",
  "author_name": "Darek Kłeczek",
  "post_date": "2021-01-27T07:04:55.019000",
  "votes": 7,
  "comment_count": 1,
  "views": 0,
  "content": "<p>From the data page: </p>\n<blockquote>\n  <p>Each sample consists of four files. Each file represents a different filter on the subcellular protein patterns represented by the sample. The format should be [filename]_[filter color].png for the PNG files. Colors are red for microtubule channels, blue for nuclei channels, yellow for Endoplasmic Reticulum (ER) channels, and green for protein of interest.</p>\n</blockquote>\n<p>I wonder what's the best way to show all 4 channels in a single image? In my <a href=\"https://www.kaggle.com/thedrcat/hpa-single-cell-classification-eda/\" target=\"_blank\">EDA kernel </a> I limited the visual representation of classes to RGB channels only, for example: <br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1328185%2Fdffda38c5cebe0cdc1fad95ce2027716%2FScreenshot%202021-01-27%20at%2008.03.36.png?generation=1611731058658510&amp;alt=media\" alt=\"cell images\"></p>",
  "messages": [
    {
      "id": 1171913,
      "postDate": "2021-01-27T07:04:55.020Z",
      "content": "<p>From the data page: </p>\n<blockquote>\n  <p>Each sample consists of four files. Each file represents a different filter on the subcellular protein patterns represented by the sample. The format should be [filename]_[filter color].png for the PNG files. Colors are red for microtubule channels, blue for nuclei channels, yellow for Endoplasmic Reticulum (ER) channels, and green for protein of interest.</p>\n</blockquote>\n<p>I wonder what's the best way to show all 4 channels in a single image? In my <a href=\"https://www.kaggle.com/thedrcat/hpa-single-cell-classification-eda/\" target=\"_blank\">EDA kernel </a> I limited the visual representation of classes to RGB channels only, for example: <br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1328185%2Fdffda38c5cebe0cdc1fad95ce2027716%2FScreenshot%202021-01-27%20at%2008.03.36.png?generation=1611731058658510&amp;alt=media\" alt=\"cell images\"></p>",
      "rawMarkdown": "From the data page: \n> Each sample consists of four files. Each file represents a different filter on the subcellular protein patterns represented by the sample. The format should be [filename]_[filter color].png for the PNG files. Colors are red for microtubule channels, blue for nuclei channels, yellow for Endoplasmic Reticulum (ER) channels, and green for protein of interest.\n\nI wonder what's the best way to show all 4 channels in a single image? In my [EDA kernel ](https://www.kaggle.com/thedrcat/hpa-single-cell-classification-eda/) I limited the visual representation of classes to RGB channels only, for example: \n![cell images](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1328185%2Fdffda38c5cebe0cdc1fad95ce2027716%2FScreenshot%202021-01-27%20at%2008.03.36.png?generation=1611731058658510&alt=media)",
      "votes": 7
    },
    {
      "id": 1172018,
      "postDate": "2021-01-27T08:11:22.307Z",
      "content": "<p>For annotation (by experts) and for the model, I guess we agree that individual channels with full range px values are better. In annotation, we toggled the channels. For visualization purpose only, you can try blending the channels. For example, red = red + yellow, green = green + yellow/2, blue=blue. </p>",
      "rawMarkdown": "For annotation (by experts) and for the model, I guess we agree that individual channels with full range px values are better. In annotation, we toggled the channels. For visualization purpose only, you can try blending the channels. For example, red = red + yellow, green = green + yellow/2, blue=blue. ",
      "votes": 4
    }
  ],
  "comments": [
    {
      "id": 1172018,
      "author_name": "Trang Le",
      "author_url": "",
      "post_date": "2021-01-27T08:11:22.307000",
      "content": "<p>For annotation (by experts) and for the model, I guess we agree that individual channels with full range px values are better. In annotation, we toggled the channels. For visualization purpose only, you can try blending the channels. For example, red = red + yellow, green = green + yellow/2, blue=blue. </p>",
      "votes": 4,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1171913": "From the data page: \n> Each sample consists of four files. Each file represents a different filter on the subcellular protein patterns represented by the sample. The format should be [filename]_[filter color].png for the PNG files. Colors are red for microtubule channels, blue for nuclei channels, yellow for Endoplasmic Reticulum (ER) channels, and green for protein of interest.\n\nI wonder what's the best way to show all 4 channels in a single image? In my [EDA kernel ](https://www.kaggle.com/thedrcat/hpa-single-cell-classification-eda/) I limited the visual representation of classes to RGB channels only, for example: \n![cell images](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1328185%2Fdffda38c5cebe0cdc1fad95ce2027716%2FScreenshot%202021-01-27%20at%2008.03.36.png?generation=1611731058658510&alt=media)",
    "1172018": "For annotation (by experts) and for the model, I guess we agree that individual channels with full range px values are better. In annotation, we toggled the channels. For visualization purpose only, you can try blending the channels. For example, red = red + yellow, green = green + yellow/2, blue=blue. "
  }
}