{
  "id": 236541,
  "title": "Labels on the test set",
  "url": "/competitions/hpa-single-cell-image-classification/discussion/236541",
  "author_name": "Rene",
  "post_date": "2021-05-04T20:49:36.709000",
  "votes": 2,
  "comment_count": 9,
  "views": 0,
  "content": "<p>Hi - just to quickly confirm: I assumed the images of the test set have unique labels, i.e. all cells in a single test picture represent the same label, is this correct?</p>\n<p>Thank you.</p>",
  "messages": [
    {
      "id": 1293418,
      "postDate": "2021-05-04T20:49:36.710Z",
      "content": "<p>Hi - just to quickly confirm: I assumed the images of the test set have unique labels, i.e. all cells in a single test picture represent the same label, is this correct?</p>\n<p>Thank you.</p>",
      "rawMarkdown": "Hi - just to quickly confirm: I assumed the images of the test set have unique labels, i.e. all cells in a single test picture represent the same label, is this correct?\n\nThank you.",
      "votes": 2
    },
    {
      "id": 1294709,
      "postDate": "2021-05-05T20:41:24.003Z",
      "content": "<p>Thank Darek you for the clarification. Do you have some specific example of the format of the submission? Looking at the sample.csv submission, it looks we need to specify a PredictionString per test picture. What would its format, e.g if  I discover a cell of type 1 and 2 in the test picture?</p>\n<p>Cheers.</p>",
      "rawMarkdown": "Thank Darek you for the clarification. Do you have some specific example of the format of the submission? Looking at the sample.csv submission, it looks we need to specify a PredictionString per test picture. What would its format, e.g if  I discover a cell of type 1 and 2 in the test picture?\n\nCheers."
    },
    {
      "id": 1294689,
      "postDate": "2021-05-05T20:25:24.773Z",
      "content": "<p>This is not correct. Actually, the images with high single cell variability were selected, so different cells are expected to represent different classes. </p>",
      "rawMarkdown": "This is not correct. Actually, the images with high single cell variability were selected, so different cells are expected to represent different classes. ",
      "replies": [
        {
          "id": 1294712,
          "postDate": "2021-05-05T20:42:46.407Z",
          "content": "<p>Thank Darek you for the clarification. Do you have some specific example of the format of the submission? Looking at the sample.csv submission, it looks we need to specify a PredictionString per test picture. What would its format, e.g if  I discover a cell of type 1 and 2 in the test picture?</p>\n<p>Cheers.</p>",
          "rawMarkdown": "Thank Darek you for the clarification. Do you have some specific example of the format of the submission? Looking at the sample.csv submission, it looks we need to specify a PredictionString per test picture. What would its format, e.g if  I discover a cell of type 1 and 2 in the test picture?\n\nCheers."
        },
        {
          "id": 1295024,
          "postDate": "2021-05-06T06:32:20.047Z",
          "content": "<p>See the evaluation page - <a href=\"https://www.kaggle.com/c/hpa-single-cell-image-classification/overview/evaluation\" target=\"_blank\">https://www.kaggle.com/c/hpa-single-cell-image-classification/overview/evaluation</a> and maybe his inference notebook will help - <a href=\"https://www.kaggle.com/thedrcat/fastai-quick-submission-template\" target=\"_blank\">https://www.kaggle.com/thedrcat/fastai-quick-submission-template</a></p>",
          "rawMarkdown": "See the evaluation page - https://www.kaggle.com/c/hpa-single-cell-image-classification/overview/evaluation and maybe his inference notebook will help - https://www.kaggle.com/thedrcat/fastai-quick-submission-template"
        },
        {
          "id": 1295229,
          "postDate": "2021-05-06T09:53:54.847Z",
          "content": "<p>Thank you, I had missed that. Reading the details I understand the following:</p>\n<ol>\n<li>In general, a test picture will have several cells, hence I need to provide several masks (In translated form), and their associated labeling.</li>\n<li>However, there seem to be situations when a particular cell (specified by its specific mask), has more than one label. This goes beyond Derek's remark above that different cells in one picture might represent different classes, now even a single cell can represent different classes.</li>\n</ol>\n<p>Does this sound right?</p>\n<p>Thank you.</p>",
          "rawMarkdown": "Thank you, I had missed that. Reading the details I understand the following:\n\n1. In general, a test picture will have several cells, hence I need to provide several masks (In translated form), and their associated labeling.\n2. However, there seem to be situations when a particular cell (specified by its specific mask), has more than one label. This goes beyond Derek's remark above that different cells in one picture might represent different classes, now even a single cell can represent different classes.\n\nDoes this sound right?\n\nThank you.\n"
        },
        {
          "id": 1295558,
          "postDate": "2021-05-06T14:52:11.177Z",
          "content": "<p>Yes both your points are correct. Each image will have multiple cells and each cell can have, and most probably will have, multiple labels. </p>",
          "rawMarkdown": "Yes both your points are correct. Each image will have multiple cells and each cell can have, and most probably will have, multiple labels. "
        },
        {
          "id": 1295564,
          "postDate": "2021-05-06T14:54:29.843Z",
          "content": "<p>Cool, thank you :)</p>",
          "rawMarkdown": "Cool, thank you :)"
        }
      ]
    },
    {
      "id": 1294154,
      "postDate": "2021-05-05T13:06:53.517Z",
      "content": "<p>Updated question - did this clarify?</p>",
      "rawMarkdown": "Updated question - did this clarify?"
    },
    {
      "id": 1293549,
      "postDate": "2021-05-05T02:21:26.017Z",
      "content": "<p>What do you mean by unique labels?</p>",
      "rawMarkdown": "What do you mean by unique labels?"
    }
  ],
  "comments": [
    {
      "id": 1294709,
      "author_name": "Rene",
      "author_url": "",
      "post_date": "2021-05-05T20:41:24.003000",
      "content": "<p>Thank Darek you for the clarification. Do you have some specific example of the format of the submission? Looking at the sample.csv submission, it looks we need to specify a PredictionString per test picture. What would its format, e.g if  I discover a cell of type 1 and 2 in the test picture?</p>\n<p>Cheers.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1294689,
      "author_name": "Darek Kłeczek",
      "author_url": "",
      "post_date": "2021-05-05T20:25:24.773000",
      "content": "<p>This is not correct. Actually, the images with high single cell variability were selected, so different cells are expected to represent different classes. </p>",
      "votes": 0,
      "replies": [
        {
          "id": 1294712,
          "author_name": "Rene",
          "author_url": "",
          "post_date": "2021-05-05T20:42:46.407000",
          "content": "<p>Thank Darek you for the clarification. Do you have some specific example of the format of the submission? Looking at the sample.csv submission, it looks we need to specify a PredictionString per test picture. What would its format, e.g if  I discover a cell of type 1 and 2 in the test picture?</p>\n<p>Cheers.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1295024,
          "author_name": "novice03",
          "author_url": "",
          "post_date": "2021-05-06T06:32:20.047000",
          "content": "<p>See the evaluation page - <a href=\"https://www.kaggle.com/c/hpa-single-cell-image-classification/overview/evaluation\" target=\"_blank\">https://www.kaggle.com/c/hpa-single-cell-image-classification/overview/evaluation</a> and maybe his inference notebook will help - <a href=\"https://www.kaggle.com/thedrcat/fastai-quick-submission-template\" target=\"_blank\">https://www.kaggle.com/thedrcat/fastai-quick-submission-template</a></p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1295229,
          "author_name": "Rene",
          "author_url": "",
          "post_date": "2021-05-06T09:53:54.847000",
          "content": "<p>Thank you, I had missed that. Reading the details I understand the following:</p>\n<ol>\n<li>In general, a test picture will have several cells, hence I need to provide several masks (In translated form), and their associated labeling.</li>\n<li>However, there seem to be situations when a particular cell (specified by its specific mask), has more than one label. This goes beyond Derek's remark above that different cells in one picture might represent different classes, now even a single cell can represent different classes.</li>\n</ol>\n<p>Does this sound right?</p>\n<p>Thank you.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1295558,
          "author_name": "novice03",
          "author_url": "",
          "post_date": "2021-05-06T14:52:11.177000",
          "content": "<p>Yes both your points are correct. Each image will have multiple cells and each cell can have, and most probably will have, multiple labels. </p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1295564,
          "author_name": "Rene",
          "author_url": "",
          "post_date": "2021-05-06T14:54:29.843000",
          "content": "<p>Cool, thank you :)</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1294154,
      "author_name": "Rene",
      "author_url": "",
      "post_date": "2021-05-05T13:06:53.517000",
      "content": "<p>Updated question - did this clarify?</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1293549,
      "author_name": "novice03",
      "author_url": "",
      "post_date": "2021-05-05T02:21:26.017000",
      "content": "<p>What do you mean by unique labels?</p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1293418": "Hi - just to quickly confirm: I assumed the images of the test set have unique labels, i.e. all cells in a single test picture represent the same label, is this correct?\n\nThank you.",
    "1294709": "Thank Darek you for the clarification. Do you have some specific example of the format of the submission? Looking at the sample.csv submission, it looks we need to specify a PredictionString per test picture. What would its format, e.g if  I discover a cell of type 1 and 2 in the test picture?\n\nCheers.",
    "1294689": "This is not correct. Actually, the images with high single cell variability were selected, so different cells are expected to represent different classes. ",
    "1294154": "Updated question - did this clarify?",
    "1293549": "What do you mean by unique labels?"
  }
}