{
  "id": 20553,
  "title": "I need Tesla!",
  "url": "/competitions/draper-satellite-image-chronology/discussion/20553",
  "author_name": "",
  "post_date": "2016-04-29T19:59:37.200Z",
  "votes": 4,
  "comment_count": 6,
  "views": 1290,
  "content": "<p>Since now a days all competitions need neural nets and deep learning, why not have a competition in which you provide a Tesla gpu to top10? ;)</p>",
  "messages": [
    {
      "id": "117602",
      "postDate": "04/29/2016 19:59:37",
      "content": "<p>Since now a days all competitions need neural nets and deep learning, why not have a competition in which you provide a Tesla gpu to top10? ;)</p>",
      "rawMarkdown": "Since now a days all competitions need neural nets and deep learning, why not have a competition in which you provide a Tesla gpu to top10? ;)",
      "votes": null
    },
    {
      "id": "117610",
      "postDate": "04/29/2016 20:31:09",
      "content": "<p>What is the basic hardware requirement for this competition? </p>",
      "rawMarkdown": "What is the basic hardware requirement for this competition?",
      "votes": null
    },
    {
      "id": "117629",
      "postDate": "04/29/2016 22:00:31",
      "content": "<p>I would not surprise if people use non-dl based method to win this competition. Understanding how to approach the problem is more important. I guess treating this problem as a simple classification problem won't work, and finetune probably not helpful, as the pretrained model are not feed with data in this visual perspective. The good part is that its high resolution, you will know what I mean when you tried hard to classify cars on low resolution street cameras</p>\n\n<p>A smart idea would beat pascal gpu I would say.</p>",
      "rawMarkdown": "I would not surprise if people use non-dl based method to win this competition. Understanding how to approach the problem is more important. I guess treating this problem as a simple classification problem won't work, and finetune probably not helpful, as the pretrained model are not feed with data in this visual perspective. The good part is that its high resolution, you will know what I mean when you tried hard to classify cars on low resolution street cameras\r\n\r\nA smart idea would beat pascal gpu I would say.",
      "votes": null
    },
    {
      "id": "117643",
      "postDate": "04/30/2016 00:18:08",
      "content": "<p>I wish, I can have a Tesla or Titan-X as well. It looks like a nice task for LSTM networks.</p>",
      "rawMarkdown": "I wish, I can have a Tesla or Titan-X as well. It looks like a nice task for LSTM networks.",
      "votes": null
    },
    {
      "id": "117644",
      "postDate": "04/30/2016 00:31:17",
      "content": "<p><img src=\"http://i.imgur.com/jI4HrID.jpg\" alt=\"enter image description here\" title></p>",
      "rawMarkdown": "![enter image description here][1]\r\n\r\n\r\n  [1]: http://i.imgur.com/jI4HrID.jpg",
      "votes": null
    },
    {
      "id": "117948",
      "postDate": "05/02/2016 00:13:19",
      "content": "<p><a href=\"http://spectrum.ieee.org/tech-talk/computing/embedded-systems/movidius-puts-neural-network-on-a-usb-stick\">How about one of these 'Fathom' thingamabobs</a></p>",
      "rawMarkdown": "[How about one of these 'Fathom' thingamabobs][1]\r\n\r\n\r\n  [1]: http://spectrum.ieee.org/tech-talk/computing/embedded-systems/movidius-puts-neural-network-on-a-usb-stick",
      "votes": null
    },
    {
      "id": "2180720",
      "postDate": "03/14/2023 03:49:08",
      "content": "<p>good idea :)</p>",
      "rawMarkdown": "good idea :)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2180720,
      "author_name": "younginit",
      "author_url": "",
      "post_date": "03/14/2023 03:49:08",
      "content": "<p>good idea :)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 117610,
      "author_name": "jialunhe",
      "author_url": "",
      "post_date": "04/29/2016 20:31:09",
      "content": "<p>What is the basic hardware requirement for this competition? </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 117629,
      "author_name": "usixuz",
      "author_url": "",
      "post_date": "04/29/2016 22:00:31",
      "content": "<p>I would not surprise if people use non-dl based method to win this competition. Understanding how to approach the problem is more important. I guess treating this problem as a simple classification problem won't work, and finetune probably not helpful, as the pretrained model are not feed with data in this visual perspective. The good part is that its high resolution, you will know what I mean when you tried hard to classify cars on low resolution street cameras</p>\n\n<p>A smart idea would beat pascal gpu I would say.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 117643,
      "author_name": "serhiy",
      "author_url": "",
      "post_date": "04/30/2016 00:18:08",
      "content": "<p>I wish, I can have a Tesla or Titan-X as well. It looks like a nice task for LSTM networks.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 117644,
      "author_name": "triskelion",
      "author_url": "",
      "post_date": "04/30/2016 00:31:17",
      "content": "<p><img src=\"http://i.imgur.com/jI4HrID.jpg\" alt=\"enter image description here\" title></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 117948,
      "author_name": "yetiman",
      "author_url": "",
      "post_date": "05/02/2016 00:13:19",
      "content": "<p><a href=\"http://spectrum.ieee.org/tech-talk/computing/embedded-systems/movidius-puts-neural-network-on-a-usb-stick\">How about one of these 'Fathom' thingamabobs</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "117602": "Since now a days all competitions need neural nets and deep learning, why not have a competition in which you provide a Tesla gpu to top10? ;)",
    "117610": "What is the basic hardware requirement for this competition?",
    "117629": "I would not surprise if people use non-dl based method to win this competition. Understanding how to approach the problem is more important. I guess treating this problem as a simple classification problem won't work, and finetune probably not helpful, as the pretrained model are not feed with data in this visual perspective. The good part is that its high resolution, you will know what I mean when you tried hard to classify cars on low resolution street cameras\r\n\r\nA smart idea would beat pascal gpu I would say.",
    "117643": "I wish, I can have a Tesla or Titan-X as well. It looks like a nice task for LSTM networks.",
    "117644": "![enter image description here][1]\r\n\r\n\r\n  [1]: http://i.imgur.com/jI4HrID.jpg",
    "117948": "[How about one of these 'Fathom' thingamabobs][1]\r\n\r\n\r\n  [1]: http://spectrum.ieee.org/tech-talk/computing/embedded-systems/movidius-puts-neural-network-on-a-usb-stick",
    "2180720": "good idea :)"
  },
  "source": "meta"
}