{
  "id": 167591,
  "title": "what  have u learnt from this competition ?( any experience)  ",
  "url": "/competitions/siim-isic-melanoma-classification/discussion/167591",
  "author_name": "",
  "post_date": "2020-07-17T07:14:34.934246200Z",
  "votes": 9,
  "comment_count": 10,
  "views": 0,
  "content": "",
  "messages": [
    {
      "id": "932612",
      "postDate": "07/17/2020 07:14:34",
      "content": "",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "933008",
      "postDate": "07/17/2020 12:22:24",
      "content": "<ul>\n<li>The importance of TTA</li>\n<li>New ways to ensemble models</li>\n</ul>",
      "rawMarkdown": "The importance of TTA\n- New ways to ensemble models",
      "votes": null
    },
    {
      "id": "933142",
      "postDate": "07/17/2020 14:18:17",
      "content": "<p>TTA,  CV, FocalLoss, creating TfRecords, integration of tabular data with CNN ... a lot</p>",
      "rawMarkdown": "TTA,  CV, FocalLoss, creating TfRecords, integration of tabular data with CNN ... a lot",
      "votes": null
    },
    {
      "id": "933154",
      "postDate": "07/17/2020 14:28:48",
      "content": "<p>i have learnt about deep learning , different architectures , focal loss ,VGG-16 model </p>",
      "rawMarkdown": "i have learnt about deep learning , different architectures , focal loss ,VGG-16 model",
      "votes": null
    },
    {
      "id": "933329",
      "postDate": "07/17/2020 16:47:01",
      "content": "<p>I've learned about:\n- Importance of data augmentations\n- TTA\n- Using TFRecords and TPU\n- Different ensembling methods\n- EfficientNet Architectures</p>",
      "rawMarkdown": "I've learned about:\n- Importance of data augmentations\n- TTA\n- Using TFRecords and TPU\n- Different ensembling methods\n- EfficientNet Architectures",
      "votes": null
    },
    {
      "id": "934251",
      "postDate": "07/18/2020 10:40:02",
      "content": "<p>What I've learnt : \n* The importance of a good CV strategy ( I used to do LB probing before)\n* Creating TF Records ( Thanks to 4x GM Chris ofc )\n* Integrating tabular and Image data into a network.\n* The importance of postprecessing (like using rankdata or poweraveraging) \n* Thinking outside the box when it comes to augmentations and being creative ( Hair removal, Hair augmentation, Microscope Aug ) </p>",
      "rawMarkdown": "What I've learnt : \n* The importance of a good CV strategy ( I used to do LB probing before)\n* Creating TF Records ( Thanks to 4x GM Chris ofc )\n* Integrating tabular and Image data into a network.\n* The importance of postprecessing (like using rankdata or poweraveraging) \n* Thinking outside the box when it comes to augmentations and being creative ( Hair removal, Hair augmentation, Microscope Aug )",
      "votes": null
    },
    {
      "id": "934583",
      "postDate": "07/18/2020 15:03:04",
      "content": "<p>You are the winner here, it doesn't even matter your final rank.\nWell done sir, guess I'll have to read more of Chris notebooks.</p>",
      "rawMarkdown": "You are the winner here, it doesn't even matter your final rank.\nWell done sir, guess I'll have to read more of Chris notebooks.",
      "votes": null
    },
    {
      "id": "934590",
      "postDate": "07/18/2020 15:05:43",
      "content": "<p>Yes you should do that, I highly recommend it they are very well explained ! </p>",
      "rawMarkdown": "Yes you should do that, I highly recommend it they are very well explained !",
      "votes": null
    },
    {
      "id": "935564",
      "postDate": "07/19/2020 13:49:01",
      "content": "<p>It's my first competition, which i try hard. Learn more about: Local Loss, data augmentation, TFRecord and importance of different imagesize   can help when processing image.</p>",
      "rawMarkdown": "It's my first competition, which i try hard. Learn more about: Local Loss, data augmentation, TFRecord and importance of different imagesize   can help when processing image.",
      "votes": null
    },
    {
      "id": "937059",
      "postDate": "07/20/2020 17:41:01",
      "content": "<p>Most of what I learned can be summed up in this article. <a href=\"https://miguel-data-sc.github.io/2017-11-23-second/\" target=\"_blank\">https://miguel-data-sc.github.io/2017-11-23-second/</a> </p>\n<p>TLDR: Size matters! </p>",
      "rawMarkdown": "Most of what I learned can be summed up in this article. https://miguel-data-sc.github.io/2017-11-23-second/ \n\nTLDR: Size matters!",
      "votes": null
    },
    {
      "id": "940816",
      "postDate": "07/23/2020 05:53:18",
      "content": "<p>Oh! This is great. Thanks.</p>",
      "rawMarkdown": "Oh! This is great. Thanks.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 937059,
      "author_name": "hooong",
      "author_url": "",
      "post_date": "07/20/2020 17:41:01",
      "content": "<p>Most of what I learned can be summed up in this article. <a href=\"https://miguel-data-sc.github.io/2017-11-23-second/\" target=\"_blank\">https://miguel-data-sc.github.io/2017-11-23-second/</a> </p>\n<p>TLDR: Size matters! </p>",
      "votes": null,
      "replies": [
        {
          "id": 940816,
          "author_name": "joshi98kishan",
          "author_url": "",
          "post_date": "07/23/2020 05:53:18",
          "content": "<p>Oh! This is great. Thanks.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 933008,
      "author_name": "abiolatti",
      "author_url": "",
      "post_date": "07/17/2020 12:22:24",
      "content": "<ul>\n<li>The importance of TTA</li>\n<li>New ways to ensemble models</li>\n</ul>",
      "votes": null,
      "replies": []
    },
    {
      "id": 933142,
      "author_name": "romanweilguny",
      "author_url": "",
      "post_date": "07/17/2020 14:18:17",
      "content": "<p>TTA,  CV, FocalLoss, creating TfRecords, integration of tabular data with CNN ... a lot</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 933154,
      "author_name": "adityakrishnavamsy",
      "author_url": "",
      "post_date": "07/17/2020 14:28:48",
      "content": "<p>i have learnt about deep learning , different architectures , focal loss ,VGG-16 model </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 933329,
      "author_name": "santiviquez",
      "author_url": "",
      "post_date": "07/17/2020 16:47:01",
      "content": "<p>I've learned about:\n- Importance of data augmentations\n- TTA\n- Using TFRecords and TPU\n- Different ensembling methods\n- EfficientNet Architectures</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 934251,
      "author_name": "aziz69",
      "author_url": "",
      "post_date": "07/18/2020 10:40:02",
      "content": "<p>What I've learnt : \n* The importance of a good CV strategy ( I used to do LB probing before)\n* Creating TF Records ( Thanks to 4x GM Chris ofc )\n* Integrating tabular and Image data into a network.\n* The importance of postprecessing (like using rankdata or poweraveraging) \n* Thinking outside the box when it comes to augmentations and being creative ( Hair removal, Hair augmentation, Microscope Aug ) </p>",
      "votes": null,
      "replies": [
        {
          "id": 934583,
          "author_name": "alincijov",
          "author_url": "",
          "post_date": "07/18/2020 15:03:04",
          "content": "<p>You are the winner here, it doesn't even matter your final rank.\nWell done sir, guess I'll have to read more of Chris notebooks.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 934590,
          "author_name": "aziz69",
          "author_url": "",
          "post_date": "07/18/2020 15:05:43",
          "content": "<p>Yes you should do that, I highly recommend it they are very well explained ! </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 935564,
      "author_name": "doanquanvietnamca",
      "author_url": "",
      "post_date": "07/19/2020 13:49:01",
      "content": "<p>It's my first competition, which i try hard. Learn more about: Local Loss, data augmentation, TFRecord and importance of different imagesize   can help when processing image.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "932612": "",
    "933008": "The importance of TTA\n- New ways to ensemble models",
    "933142": "TTA,  CV, FocalLoss, creating TfRecords, integration of tabular data with CNN ... a lot",
    "933154": "i have learnt about deep learning , different architectures , focal loss ,VGG-16 model",
    "933329": "I've learned about:\n- Importance of data augmentations\n- TTA\n- Using TFRecords and TPU\n- Different ensembling methods\n- EfficientNet Architectures",
    "934251": "What I've learnt : \n* The importance of a good CV strategy ( I used to do LB probing before)\n* Creating TF Records ( Thanks to 4x GM Chris ofc )\n* Integrating tabular and Image data into a network.\n* The importance of postprecessing (like using rankdata or poweraveraging) \n* Thinking outside the box when it comes to augmentations and being creative ( Hair removal, Hair augmentation, Microscope Aug )",
    "934583": "You are the winner here, it doesn't even matter your final rank.\nWell done sir, guess I'll have to read more of Chris notebooks.",
    "934590": "Yes you should do that, I highly recommend it they are very well explained !",
    "935564": "It's my first competition, which i try hard. Learn more about: Local Loss, data augmentation, TFRecord and importance of different imagesize   can help when processing image.",
    "937059": "Most of what I learned can be summed up in this article. https://miguel-data-sc.github.io/2017-11-23-second/ \n\nTLDR: Size matters!",
    "940816": "Oh! This is great. Thanks."
  },
  "source": "meta"
}