{
  "id": 76881,
  "title": "HPA image 3 channels",
  "url": "/competitions/human-protein-atlas-image-classification/discussion/76881",
  "author_name": "",
  "post_date": "2019-01-07T14:25:44.447053600Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>HPA images I have downloaded have 3 channels for each of R, G ,B, and Y,  but the images in the original dataset have only one. So when using these images what strategy is to follow. Do we convert the images to just grayscale and then use them ?\nI used the data from this link.\n<a href=\"https://www.kaggle.com/therealpythonman/get-350k-additional-hpa-images\">https://www.kaggle.com/therealpythonman/get-350k-additional-hpa-images</a></p>",
  "messages": [
    {
      "id": "451709",
      "postDate": "01/07/2019 14:25:44",
      "content": "<p>HPA images I have downloaded have 3 channels for each of R, G ,B, and Y,  but the images in the original dataset have only one. So when using these images what strategy is to follow. Do we convert the images to just grayscale and then use them ?\nI used the data from this link.\n<a href=\"https://www.kaggle.com/therealpythonman/get-350k-additional-hpa-images\">https://www.kaggle.com/therealpythonman/get-350k-additional-hpa-images</a></p>",
      "rawMarkdown": "HPA images I have downloaded have 3 channels for each of R, G ,B, and Y,  but the images in the original dataset have only one. So when using these images what strategy is to follow. Do we convert the images to just grayscale and then use them ?\nI used the data from this link.\n[https://www.kaggle.com/therealpythonman/get-350k-additional-hpa-images][1]\n\n\n  [1]: https://www.kaggle.com/therealpythonman/get-350k-additional-hpa-images",
      "votes": null
    },
    {
      "id": "453213",
      "postDate": "01/09/2019 21:22:04",
      "content": "<p>No, please dont convert them to grayscale images. As far as i know, these additional channels per image is noise that's been added to color the images. For red image, use red[ : , :, 0 ], for green channel use green[ : , :, 1] and for blue use blue[ :, :, 2], and for yellow you can use either the red or green channel of the yellow image.(that's yellow[ :, :, 0 or 1]). Thanks</p>",
      "rawMarkdown": "No, please dont convert them to grayscale images. As far as i know, these additional channels per image is noise that's been added to color the images. For red image, use red[ : , :, 0 ], for green channel use green[ : , :, 1] and for blue use blue[ :, :, 2], and for yellow you can use either the red or green channel of the yellow image.(that's yellow[ :, :, 0 or 1]). Thanks",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 453213,
      "author_name": "adilurrahim",
      "author_url": "",
      "post_date": "01/09/2019 21:22:04",
      "content": "<p>No, please dont convert them to grayscale images. As far as i know, these additional channels per image is noise that's been added to color the images. For red image, use red[ : , :, 0 ], for green channel use green[ : , :, 1] and for blue use blue[ :, :, 2], and for yellow you can use either the red or green channel of the yellow image.(that's yellow[ :, :, 0 or 1]). Thanks</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "451709": "HPA images I have downloaded have 3 channels for each of R, G ,B, and Y,  but the images in the original dataset have only one. So when using these images what strategy is to follow. Do we convert the images to just grayscale and then use them ?\nI used the data from this link.\n[https://www.kaggle.com/therealpythonman/get-350k-additional-hpa-images][1]\n\n\n  [1]: https://www.kaggle.com/therealpythonman/get-350k-additional-hpa-images",
    "453213": "No, please dont convert them to grayscale images. As far as i know, these additional channels per image is noise that's been added to color the images. For red image, use red[ : , :, 0 ], for green channel use green[ : , :, 1] and for blue use blue[ :, :, 2], and for yellow you can use either the red or green channel of the yellow image.(that's yellow[ :, :, 0 or 1]). Thanks"
  },
  "source": "meta"
}