{
  "id": 571212,
  "title": "What score would be useful in practice?",
  "url": "/competitions/byu-locating-bacterial-flagellar-motors-2025/discussion/571212",
  "author_name": "",
  "post_date": "2025-04-01T22:24:59.263911100Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Just curious, how high of a score a model has to get so that you can actually use it to eliminate human work?</p>",
  "messages": [
    {
      "id": "3167869",
      "postDate": "04/01/2025 22:24:59",
      "content": "<p>Just curious, how high of a score a model has to get so that you can actually use it to eliminate human work?</p>",
      "rawMarkdown": "Just curious, how high of a score a model has to get so that you can actually use it to eliminate human work?",
      "votes": null
    },
    {
      "id": "3168101",
      "postDate": "04/02/2025 05:54:35",
      "content": "<p>Depends on what your definition of eliminating human work is. </p>\n<p>If our only goal is to provide a fast solution to replace template matching, then the top leaderboard scores are likely enough. </p>\n<p>If our goal is to locate some quantity of flagellar motors so we can do sub-tomogram averaging, then the baseline models provided in the sample notebooks are sufficient. </p>\n<p>If our goal is to find every single flagellar motor for sub-tomogram averaging, I don't really think that is possible. </p>\n<p>The goal of this competition is twofold. First, to come up with a method that beats template matching and we can apply to other macro-molecular structures. I'd say we attained this goal within the first two weeks of this competition. Second, we want to jumpstart our progress to the holy grail of making a model that is able to detect anything unusual and find things that the human eye would miss. It really remains to be seen exactly how the work here would contribute to this goal, but I'm very optimistic as I sign on every day and see the cross-pollination of ideas that I'd never even conceived!</p>",
      "rawMarkdown": "Depends on what your definition of eliminating human work is. \n\nIf our only goal is to provide a fast solution to replace template matching, then the top leaderboard scores are likely enough. \n\nIf our goal is to locate some quantity of flagellar motors so we can do sub-tomogram averaging, then the baseline models provided in the sample notebooks are sufficient. \n\nIf our goal is to find every single flagellar motor for sub-tomogram averaging, I don't really think that is possible. \n\nThe goal of this competition is twofold. First, to come up with a method that beats template matching and we can apply to other macro-molecular structures. I'd say we attained this goal within the first two weeks of this competition. Second, we want to jumpstart our progress to the holy grail of making a model that is able to detect anything unusual and find things that the human eye would miss. It really remains to be seen exactly how the work here would contribute to this goal, but I'm very optimistic as I sign on every day and see the cross-pollination of ideas that I'd never even conceived!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3168101,
      "author_name": "andrewjdarley",
      "author_url": "",
      "post_date": "04/02/2025 05:54:35",
      "content": "<p>Depends on what your definition of eliminating human work is. </p>\n<p>If our only goal is to provide a fast solution to replace template matching, then the top leaderboard scores are likely enough. </p>\n<p>If our goal is to locate some quantity of flagellar motors so we can do sub-tomogram averaging, then the baseline models provided in the sample notebooks are sufficient. </p>\n<p>If our goal is to find every single flagellar motor for sub-tomogram averaging, I don't really think that is possible. </p>\n<p>The goal of this competition is twofold. First, to come up with a method that beats template matching and we can apply to other macro-molecular structures. I'd say we attained this goal within the first two weeks of this competition. Second, we want to jumpstart our progress to the holy grail of making a model that is able to detect anything unusual and find things that the human eye would miss. It really remains to be seen exactly how the work here would contribute to this goal, but I'm very optimistic as I sign on every day and see the cross-pollination of ideas that I'd never even conceived!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3167869": "Just curious, how high of a score a model has to get so that you can actually use it to eliminate human work?",
    "3168101": "Depends on what your definition of eliminating human work is. \n\nIf our only goal is to provide a fast solution to replace template matching, then the top leaderboard scores are likely enough. \n\nIf our goal is to locate some quantity of flagellar motors so we can do sub-tomogram averaging, then the baseline models provided in the sample notebooks are sufficient. \n\nIf our goal is to find every single flagellar motor for sub-tomogram averaging, I don't really think that is possible. \n\nThe goal of this competition is twofold. First, to come up with a method that beats template matching and we can apply to other macro-molecular structures. I'd say we attained this goal within the first two weeks of this competition. Second, we want to jumpstart our progress to the holy grail of making a model that is able to detect anything unusual and find things that the human eye would miss. It really remains to be seen exactly how the work here would contribute to this goal, but I'm very optimistic as I sign on every day and see the cross-pollination of ideas that I'd never even conceived!"
  },
  "source": "meta"
}