{
  "id": 289030,
  "title": "Some questions in the `LIVECell_dataset_2021` datasets",
  "url": "/competitions/sartorius-cell-instance-segmentation/discussion/289030",
  "author_name": "",
  "post_date": "2021-11-19T02:26:20.050430700Z",
  "votes": 7,
  "comment_count": 5,
  "views": 0,
  "content": "<p>In this comp i found that <code>LIVECell_dataset_2021</code> seemed never used, i don't know how to get good use of it. Do you guys have some ideals?</p>",
  "messages": [
    {
      "id": "1587862",
      "postDate": "11/19/2021 02:26:20",
      "content": "<p>In this comp i found that <code>LIVECell_dataset_2021</code> seemed never used, i don't know how to get good use of it. Do you guys have some ideals?</p>",
      "rawMarkdown": "In this comp i found that `LIVECell_dataset_2021` seemed never used, i don't know how to get good use of it. Do you guys have some ideals?",
      "votes": null
    },
    {
      "id": "1587964",
      "postDate": "11/19/2021 04:58:17",
      "content": "<p>Read LIVECell paper for the inspiration<br>\n<a href=\"https://www.nature.com/articles/s41592-021-01249-6.pdf\" target=\"_blank\">https://www.nature.com/articles/s41592-021-01249-6.pdf</a></p>",
      "rawMarkdown": "Read LIVECell paper for the inspiration\nhttps://www.nature.com/articles/s41592-021-01249-6.pdf",
      "votes": null
    },
    {
      "id": "1587998",
      "postDate": "11/19/2021 05:51:11",
      "content": "<p>Thx a lot!</p>",
      "rawMarkdown": "Thx a lot!",
      "votes": null
    },
    {
      "id": "1588714",
      "postDate": "11/19/2021 15:19:35",
      "content": "<p>You can think of it like a dataset for making a pre-trained network for transfer learning. More data is always better, you just may need to get creative 👍😃</p>",
      "rawMarkdown": "You can think of it like a dataset for making a pre-trained network for transfer learning. More data is always better, you just may need to get creative 👍😃",
      "votes": null
    },
    {
      "id": "1589271",
      "postDate": "11/20/2021 05:06:46",
      "content": "<p>I got it, thanks a lot</p>",
      "rawMarkdown": "I got it, thanks a lot",
      "votes": null
    },
    {
      "id": "1590764",
      "postDate": "11/21/2021 16:24:09",
      "content": "<p>Train a model using the train data. Predict the livecell_dataset_2021 instances. Taking a look at the predictions'  confidences, identify some good samples from the livecell_dataset_2021 and add them to the train data. Train a new model with this final dataset.</p>\n<p>The trick is you need to evaluate the metric in the second model using only the original train data. If the metric is better than that of the first trained model's (using only the original train data), your second model performs better.</p>",
      "rawMarkdown": "Train a model using the train data. Predict the livecell_dataset_2021 instances. Taking a look at the predictions'  confidences, identify some good samples from the livecell_dataset_2021 and add them to the train data. Train a new model with this final dataset.\n\nThe trick is you need to evaluate the metric in the second model using only the original train data. If the metric is better than that of the first trained model's (using only the original train data), your second model performs better.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1587964,
      "author_name": "gunesevitan",
      "author_url": "",
      "post_date": "11/19/2021 04:58:17",
      "content": "<p>Read LIVECell paper for the inspiration<br>\n<a href=\"https://www.nature.com/articles/s41592-021-01249-6.pdf\" target=\"_blank\">https://www.nature.com/articles/s41592-021-01249-6.pdf</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 1587998,
          "author_name": "manyuli",
          "author_url": "",
          "post_date": "11/19/2021 05:51:11",
          "content": "<p>Thx a lot!</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1588714,
      "author_name": "wsmonroe",
      "author_url": "",
      "post_date": "11/19/2021 15:19:35",
      "content": "<p>You can think of it like a dataset for making a pre-trained network for transfer learning. More data is always better, you just may need to get creative 👍😃</p>",
      "votes": null,
      "replies": [
        {
          "id": 1589271,
          "author_name": "manyuli",
          "author_url": "",
          "post_date": "11/20/2021 05:06:46",
          "content": "<p>I got it, thanks a lot</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1590764,
      "author_name": "tolgadincer",
      "author_url": "",
      "post_date": "11/21/2021 16:24:09",
      "content": "<p>Train a model using the train data. Predict the livecell_dataset_2021 instances. Taking a look at the predictions'  confidences, identify some good samples from the livecell_dataset_2021 and add them to the train data. Train a new model with this final dataset.</p>\n<p>The trick is you need to evaluate the metric in the second model using only the original train data. If the metric is better than that of the first trained model's (using only the original train data), your second model performs better.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1587862": "In this comp i found that `LIVECell_dataset_2021` seemed never used, i don't know how to get good use of it. Do you guys have some ideals?",
    "1587964": "Read LIVECell paper for the inspiration\nhttps://www.nature.com/articles/s41592-021-01249-6.pdf",
    "1587998": "Thx a lot!",
    "1588714": "You can think of it like a dataset for making a pre-trained network for transfer learning. More data is always better, you just may need to get creative 👍😃",
    "1589271": "I got it, thanks a lot",
    "1590764": "Train a model using the train data. Predict the livecell_dataset_2021 instances. Taking a look at the predictions'  confidences, identify some good samples from the livecell_dataset_2021 and add them to the train data. Train a new model with this final dataset.\n\nThe trick is you need to evaluate the metric in the second model using only the original train data. If the metric is better than that of the first trained model's (using only the original train data), your second model performs better."
  },
  "source": "meta"
}