{
  "id": 98495,
  "title": "How would you train on the massive dataset?",
  "url": "/competitions/open-images-2019-object-detection/discussion/98495",
  "author_name": "",
  "post_date": "2019-07-04T06:51:22.035781700Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi, </p>\n\n<p>I'm just wondering how you guys would train a model on the massive 1.7M images. My impressions from the last year competition is that I probably need a cluster of GPUs, but no way to access to the resources for me. For me even the fine-tuning would require tens of thousands of images and GPUs. Any tips would be appreciated!</p>",
  "messages": [
    {
      "id": "567918",
      "postDate": "07/04/2019 06:51:22",
      "content": "<p>Hi, </p>\n\n<p>I'm just wondering how you guys would train a model on the massive 1.7M images. My impressions from the last year competition is that I probably need a cluster of GPUs, but no way to access to the resources for me. For me even the fine-tuning would require tens of thousands of images and GPUs. Any tips would be appreciated!</p>",
      "rawMarkdown": "Hi, \n\nI'm just wondering how you guys would train a model on the massive 1.7M images. My impressions from the last year competition is that I probably need a cluster of GPUs, but no way to access to the resources for me. For me even the fine-tuning would require tens of thousands of images and GPUs. Any tips would be appreciated!",
      "votes": null
    },
    {
      "id": "567960",
      "postDate": "07/04/2019 07:37:19",
      "content": "<p>try GCP , maybe that will help </p>",
      "rawMarkdown": "try GCP , maybe that will help",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 567960,
      "author_name": "iluvmahheart",
      "author_url": "",
      "post_date": "07/04/2019 07:37:19",
      "content": "<p>try GCP , maybe that will help </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "567918": "Hi, \n\nI'm just wondering how you guys would train a model on the massive 1.7M images. My impressions from the last year competition is that I probably need a cluster of GPUs, but no way to access to the resources for me. For me even the fine-tuning would require tens of thousands of images and GPUs. Any tips would be appreciated!",
    "567960": "try GCP , maybe that will help"
  },
  "source": "meta"
}