{
  "id": 236623,
  "title": "Using HPACellSeg in final submissions",
  "url": "/competitions/hpa-single-cell-image-classification/discussion/236623",
  "author_name": "novice03",
  "post_date": "2021-05-05T05:25:53.787000",
  "votes": 6,
  "comment_count": 10,
  "views": 0,
  "content": "<p>If my understanding is correct, there will be close to 1200 images in the test set. The unmodified HPACellSeg seems to be taking no more than 1.5 hours for about 500 images. At this rate, getting masks for all private test images should be done in about 4 to 5 hours. This seems to be good enough considering that there are a few hours left for inference. So, is it safe to use the HPACellSeg unmodified when making the final submissions or is it best to use a faster version just in case? What is everyone else doing for their private test set submissions?</p>",
  "messages": [
    {
      "id": 1293699,
      "postDate": "2021-05-05T05:25:53.787Z",
      "content": "<p>If my understanding is correct, there will be close to 1200 images in the test set. The unmodified HPACellSeg seems to be taking no more than 1.5 hours for about 500 images. At this rate, getting masks for all private test images should be done in about 4 to 5 hours. This seems to be good enough considering that there are a few hours left for inference. So, is it safe to use the HPACellSeg unmodified when making the final submissions or is it best to use a faster version just in case? What is everyone else doing for their private test set submissions?</p>",
      "rawMarkdown": "If my understanding is correct, there will be close to 1200 images in the test set. The unmodified HPACellSeg seems to be taking no more than 1.5 hours for about 500 images. At this rate, getting masks for all private test images should be done in about 4 to 5 hours. This seems to be good enough considering that there are a few hours left for inference. So, is it safe to use the HPACellSeg unmodified when making the final submissions or is it best to use a faster version just in case? What is everyone else doing for their private test set submissions?",
      "votes": 6
    },
    {
      "id": 1293760,
      "postDate": "2021-05-05T06:33:03.810Z",
      "content": "<p>When I set Private = False , It will exceed the time.</p>",
      "rawMarkdown": "When I set Private = False , It will exceed the time.",
      "votes": 2,
      "replies": [
        {
          "id": 1293798,
          "postDate": "2021-05-05T07:08:11.430Z",
          "content": "<p>You mean it takes more than 9 hours to get segmentation of 559 test images?</p>",
          "rawMarkdown": "You mean it takes more than 9 hours to get segmentation of 559 test images?"
        },
        {
          "id": 1293821,
          "postDate": "2021-05-05T07:25:04.117Z",
          "content": "<p>Yes.Or it will exceed the computation.</p>",
          "rawMarkdown": "Yes.Or it will exceed the computation."
        },
        {
          "id": 1293829,
          "postDate": "2021-05-05T07:34:03.700Z",
          "content": "<p>That's strange. How exactly are you getting the masks? Are you using an approach similar to <a href=\"https://www.kaggle.com/lnhtrang/hpa-public-data-download-and-hpacellseg\" target=\"_blank\">this</a> notebook? I completed segmentation of 1500 images in 3.5 hours just now with the original model.</p>",
          "rawMarkdown": "That's strange. How exactly are you getting the masks? Are you using an approach similar to [this](https://www.kaggle.com/lnhtrang/hpa-public-data-download-and-hpacellseg) notebook? I completed segmentation of 1500 images in 3.5 hours just now with the original model."
        },
        {
          "id": 1293854,
          "postDate": "2021-05-05T08:02:49.820Z",
          "content": "<p>I run this [(url)<a href=\"https://www.kaggle.com/jitshil143/human-protein-atlas\" target=\"_blank\">https://www.kaggle.com/jitshil143/human-protein-atlas</a>]</p>",
          "rawMarkdown": "I run this [(url)https://www.kaggle.com/jitshil143/human-protein-atlas]",
          "votes": 2
        },
        {
          "id": 1293859,
          "postDate": "2021-05-05T08:05:43.770Z",
          "content": "<p>Hi <a href=\"https://www.kaggle.com/zekunn\" target=\"_blank\">@zekunn</a> </p>\n<p>Does this notebook work on the private LB? Its set to public only right.</p>",
          "rawMarkdown": "Hi @zekunn \n\nDoes this notebook work on the private LB? Its set to public only right.",
          "votes": 1
        },
        {
          "id": 1293864,
          "postDate": "2021-05-05T08:10:14.170Z",
          "content": "<p>Just set the param 'isPublic' to False.It will cause the problem.</p>",
          "rawMarkdown": "Just set the param 'isPublic' to False.It will cause the problem.",
          "votes": 1
        }
      ]
    },
    {
      "id": 1293846,
      "postDate": "2021-05-05T07:53:48.247Z",
      "content": "<p>How are you submitting? Are you extracting mask first then inference or in the time of extracting mask you are calculating the probabilities and all.</p>",
      "rawMarkdown": "How are you submitting? Are you extracting mask first then inference or in the time of extracting mask you are calculating the probabilities and all.",
      "replies": [
        {
          "id": 1293953,
          "postDate": "2021-05-05T09:52:05.717Z",
          "content": "<p>My model needs images of single cells. Right now, I am getting all the single cells in the test set, and then running my model on those images.</p>",
          "rawMarkdown": "My model needs images of single cells. Right now, I am getting all the single cells in the test set, and then running my model on those images."
        },
        {
          "id": 1294232,
          "postDate": "2021-05-05T13:46:05.243Z",
          "content": "<p>Can you please share sample notebook for submitting. I am facing problems on submitting final prediction. Please help.🙏</p>",
          "rawMarkdown": "Can you please share sample notebook for submitting. I am facing problems on submitting final prediction. Please help.🙏"
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 1293760,
      "author_name": "Zekun",
      "author_url": "",
      "post_date": "2021-05-05T06:33:03.810000",
      "content": "<p>When I set Private = False , It will exceed the time.</p>",
      "votes": 2,
      "replies": [
        {
          "id": 1293798,
          "author_name": "novice03",
          "author_url": "",
          "post_date": "2021-05-05T07:08:11.430000",
          "content": "<p>You mean it takes more than 9 hours to get segmentation of 559 test images?</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1293821,
          "author_name": "Zekun",
          "author_url": "",
          "post_date": "2021-05-05T07:25:04.117000",
          "content": "<p>Yes.Or it will exceed the computation.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1293829,
          "author_name": "novice03",
          "author_url": "",
          "post_date": "2021-05-05T07:34:03.700000",
          "content": "<p>That's strange. How exactly are you getting the masks? Are you using an approach similar to <a href=\"https://www.kaggle.com/lnhtrang/hpa-public-data-download-and-hpacellseg\" target=\"_blank\">this</a> notebook? I completed segmentation of 1500 images in 3.5 hours just now with the original model.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1293854,
          "author_name": "Zekun",
          "author_url": "",
          "post_date": "2021-05-05T08:02:49.820000",
          "content": "<p>I run this [(url)<a href=\"https://www.kaggle.com/jitshil143/human-protein-atlas\" target=\"_blank\">https://www.kaggle.com/jitshil143/human-protein-atlas</a>]</p>",
          "votes": 2,
          "replies": []
        },
        {
          "id": 1293859,
          "author_name": "Jagadish Sivakumar",
          "author_url": "",
          "post_date": "2021-05-05T08:05:43.770000",
          "content": "<p>Hi <a href=\"https://www.kaggle.com/zekunn\" target=\"_blank\">@zekunn</a> </p>\n<p>Does this notebook work on the private LB? Its set to public only right.</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 1293864,
          "author_name": "Zekun",
          "author_url": "",
          "post_date": "2021-05-05T08:10:14.170000",
          "content": "<p>Just set the param 'isPublic' to False.It will cause the problem.</p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 1293846,
      "author_name": "badmiton",
      "author_url": "",
      "post_date": "2021-05-05T07:53:48.247000",
      "content": "<p>How are you submitting? Are you extracting mask first then inference or in the time of extracting mask you are calculating the probabilities and all.</p>",
      "votes": 0,
      "replies": [
        {
          "id": 1293953,
          "author_name": "novice03",
          "author_url": "",
          "post_date": "2021-05-05T09:52:05.717000",
          "content": "<p>My model needs images of single cells. Right now, I am getting all the single cells in the test set, and then running my model on those images.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1294232,
          "author_name": "badmiton",
          "author_url": "",
          "post_date": "2021-05-05T13:46:05.243000",
          "content": "<p>Can you please share sample notebook for submitting. I am facing problems on submitting final prediction. Please help.🙏</p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1293699": "If my understanding is correct, there will be close to 1200 images in the test set. The unmodified HPACellSeg seems to be taking no more than 1.5 hours for about 500 images. At this rate, getting masks for all private test images should be done in about 4 to 5 hours. This seems to be good enough considering that there are a few hours left for inference. So, is it safe to use the HPACellSeg unmodified when making the final submissions or is it best to use a faster version just in case? What is everyone else doing for their private test set submissions?",
    "1293760": "When I set Private = False , It will exceed the time.",
    "1293846": "How are you submitting? Are you extracting mask first then inference or in the time of extracting mask you are calculating the probabilities and all."
  }
}