{
  "id": 307946,
  "title": "Why the Eff (net) ?",
  "url": "/competitions/happy-whale-and-dolphin/discussion/307946",
  "author_name": "",
  "post_date": "2022-02-16T11:07:27.610534800Z",
  "votes": 10,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Why is everyone using efficient net in the public notebooks ?!?!?!?!?!?!?!?!</p>",
  "messages": [
    {
      "id": "1692944",
      "postDate": "02/16/2022 11:07:27",
      "content": "<p>Why is everyone using efficient net in the public notebooks ?!?!?!?!?!?!?!?!</p>",
      "rawMarkdown": "Why is everyone using efficient net in the public notebooks ?!?!?!?!?!?!?!?!",
      "votes": null
    },
    {
      "id": "1693043",
      "postDate": "02/16/2022 12:08:49",
      "content": "<p>Since the Efficientnet is light compared to other networks without degrading performance</p>",
      "rawMarkdown": "Since the Efficientnet is light compared to other networks without degrading performance",
      "votes": null
    },
    {
      "id": "1693057",
      "postDate": "02/16/2022 12:13:02",
      "content": "<p>Because many of the notebooks are basically copying and pasting the first original high scoring notebook. They just change image size or something and get a better score ;)</p>",
      "rawMarkdown": "Because many of the notebooks are basically copying and pasting the first original high scoring notebook. They just change image size or something and get a better score ;)",
      "votes": null
    },
    {
      "id": "1693451",
      "postDate": "02/16/2022 17:14:31",
      "content": "<p>Efficientnet architecture models are a good start for a baseline. There are fast and perform pretty good in almost every task. You can easily scale the size (B0 through B7) and resolution to find a sweet spot. <br>\nAs the competition progress, more and more architectures and methodologies will be developed and compared</p>",
      "rawMarkdown": "Efficientnet architecture models are a good start for a baseline. There are fast and perform pretty good in almost every task. You can easily scale the size (B0 through B7) and resolution to find a sweet spot. \nAs the competition progress, more and more architectures and methodologies will be developed and compared",
      "votes": null
    },
    {
      "id": "1693467",
      "postDate": "02/16/2022 17:28:33",
      "content": "<p>It's also super early in the competition! Usually a few weeks into the competition, we start seeing interesting bits being shared. </p>",
      "rawMarkdown": "It's also super early in the competition! Usually a few weeks into the competition, we start seeing interesting bits being shared.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1693043,
      "author_name": "deepkim",
      "author_url": "",
      "post_date": "02/16/2022 12:08:49",
      "content": "<p>Since the Efficientnet is light compared to other networks without degrading performance</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1693057,
      "author_name": "taindow",
      "author_url": "",
      "post_date": "02/16/2022 12:13:02",
      "content": "<p>Because many of the notebooks are basically copying and pasting the first original high scoring notebook. They just change image size or something and get a better score ;)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1693451,
      "author_name": "vladvdv",
      "author_url": "",
      "post_date": "02/16/2022 17:14:31",
      "content": "<p>Efficientnet architecture models are a good start for a baseline. There are fast and perform pretty good in almost every task. You can easily scale the size (B0 through B7) and resolution to find a sweet spot. <br>\nAs the competition progress, more and more architectures and methodologies will be developed and compared</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1693467,
      "author_name": "init27",
      "author_url": "",
      "post_date": "02/16/2022 17:28:33",
      "content": "<p>It's also super early in the competition! Usually a few weeks into the competition, we start seeing interesting bits being shared. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1692944": "Why is everyone using efficient net in the public notebooks ?!?!?!?!?!?!?!?!",
    "1693043": "Since the Efficientnet is light compared to other networks without degrading performance",
    "1693057": "Because many of the notebooks are basically copying and pasting the first original high scoring notebook. They just change image size or something and get a better score ;)",
    "1693451": "Efficientnet architecture models are a good start for a baseline. There are fast and perform pretty good in almost every task. You can easily scale the size (B0 through B7) and resolution to find a sweet spot. \nAs the competition progress, more and more architectures and methodologies will be developed and compared",
    "1693467": "It's also super early in the competition! Usually a few weeks into the competition, we start seeing interesting bits being shared."
  },
  "source": "meta"
}