{
  "id": 72979,
  "title": "what is the best stroke strategy?",
  "url": "/competitions/quickdraw-doodle-recognition/discussion/72979",
  "author_name": "remidi",
  "post_date": "2018-11-28T19:32:21.631000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I am using simple single channel, plain simple lines with resnet50 for 0.931 LB. \nStrategies till now :</p>\n\n<ul>\n<li>@beluga -- encoding <strong>time</strong> of stroke in single channel with color intensity variance</li>\n<li>@Appian and @Heng -- encoding <strong>3 channel</strong> , with stroke velocity, and line (train_raw: 65gb??)</li>\n<li>@Omallo -- using <strong>6 channels / 3 channels</strong> with percentage of strokes in each channel</li>\n</ul>\n\n<p><strong>what is the best strategy, and are there any references for the strategies ?</strong></p>\n\n<p>Reference: </p>\n\n<ol>\n<li><a href=\"https://www.kaggle.com/c/quickdraw-doodle-recognition/discussion/70558#424280\">https://www.kaggle.com/c/quickdraw-doodle-recognition/discussion/70558#424280</a></li>\n<li><a href=\"https://www.kaggle.com/gaborfodor/greyscale-mobilenet-lb-0-892\">https://www.kaggle.com/gaborfodor/greyscale-mobilenet-lb-0-892</a></li>\n</ol>",
  "messages": [
    {
      "id": 429388,
      "postDate": "2018-11-28T19:32:21.633Z",
      "content": "<p>I am using simple single channel, plain simple lines with resnet50 for 0.931 LB. \nStrategies till now :</p>\n\n<ul>\n<li>@beluga -- encoding <strong>time</strong> of stroke in single channel with color intensity variance</li>\n<li>@Appian and @Heng -- encoding <strong>3 channel</strong> , with stroke velocity, and line (train_raw: 65gb??)</li>\n<li>@Omallo -- using <strong>6 channels / 3 channels</strong> with percentage of strokes in each channel</li>\n</ul>\n\n<p><strong>what is the best strategy, and are there any references for the strategies ?</strong></p>\n\n<p>Reference: </p>\n\n<ol>\n<li><a href=\"https://www.kaggle.com/c/quickdraw-doodle-recognition/discussion/70558#424280\">https://www.kaggle.com/c/quickdraw-doodle-recognition/discussion/70558#424280</a></li>\n<li><a href=\"https://www.kaggle.com/gaborfodor/greyscale-mobilenet-lb-0-892\">https://www.kaggle.com/gaborfodor/greyscale-mobilenet-lb-0-892</a></li>\n</ol>",
      "rawMarkdown": "I am using simple single channel, plain simple lines with resnet50 for 0.931 LB. \nStrategies till now :\n\n+ @beluga -- encoding **time** of stroke in single channel with color intensity variance\n+ @Appian and @Heng -- encoding **3 channel** , with stroke velocity, and line (train_raw: 65gb??)\n+ @Omallo -- using **6 channels / 3 channels** with percentage of strokes in each channel\n\n**what is the best strategy, and are there any references for the strategies ?**\n\n\nReference: \n\n 1. https://www.kaggle.com/c/quickdraw-doodle-recognition/discussion/70558#424280\n 2. https://www.kaggle.com/gaborfodor/greyscale-mobilenet-lb-0-892"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "429388": "I am using simple single channel, plain simple lines with resnet50 for 0.931 LB. \nStrategies till now :\n\n+ @beluga -- encoding **time** of stroke in single channel with color intensity variance\n+ @Appian and @Heng -- encoding **3 channel** , with stroke velocity, and line (train_raw: 65gb??)\n+ @Omallo -- using **6 channels / 3 channels** with percentage of strokes in each channel\n\n**what is the best strategy, and are there any references for the strategies ?**\n\n\nReference: \n\n 1. https://www.kaggle.com/c/quickdraw-doodle-recognition/discussion/70558#424280\n 2. https://www.kaggle.com/gaborfodor/greyscale-mobilenet-lb-0-892"
  }
}