{
  "id": 91386,
  "title": "Keras MixUp on Preprocessed Mel-Spectrogram Data [LB632]",
  "url": "/competitions/freesound-audio-tagging-2019/discussion/91386",
  "author_name": "",
  "post_date": "2019-05-04T03:45:38.691147500Z",
  "votes": 4,
  "comment_count": 4,
  "views": 0,
  "content": "<p><strong>EDIT: Now, with a handful of augmentations, the kernel finally achieve 0.632 LB score  ... hope it is helpful!</strong></p>\n\n<p>Hi everyone!</p>\n\n<p>First of all, thanks <a href=\"/daisukelab\">@daisukelab</a> for his great kernel on preprocessed Mel-Spectrogram data. This data very much benefit all of participants here. However, currently only Pytorch-kernel implementation (such as one by <a href=\"/mhiro2\">@mhiro2</a> ) exists. </p>\n\n<p>If anyone is like me who also would like to try this data based on Keras implementation, you can have a Keras starter kernel here : </p>\n\n<p><a href=\"https://www.kaggle.com/ratthachat/fat19-kerasbaseline-on-preprocesseddata-lb576/\">https://www.kaggle.com/ratthachat/fat19-kerasbaseline-on-preprocesseddata-lb576/</a></p>\n\n<p>Hope it is helpful and happy kaggling everybody!</p>",
  "messages": [
    {
      "id": "526897",
      "postDate": "05/04/2019 03:45:38",
      "content": "<p><strong>EDIT: Now, with a handful of augmentations, the kernel finally achieve 0.632 LB score  ... hope it is helpful!</strong></p>\n\n<p>Hi everyone!</p>\n\n<p>First of all, thanks <a href=\"/daisukelab\">@daisukelab</a> for his great kernel on preprocessed Mel-Spectrogram data. This data very much benefit all of participants here. However, currently only Pytorch-kernel implementation (such as one by <a href=\"/mhiro2\">@mhiro2</a> ) exists. </p>\n\n<p>If anyone is like me who also would like to try this data based on Keras implementation, you can have a Keras starter kernel here : </p>\n\n<p><a href=\"https://www.kaggle.com/ratthachat/fat19-kerasbaseline-on-preprocesseddata-lb576/\">https://www.kaggle.com/ratthachat/fat19-kerasbaseline-on-preprocesseddata-lb576/</a></p>\n\n<p>Hope it is helpful and happy kaggling everybody!</p>",
      "rawMarkdown": "**EDIT: Now, with a handful of augmentations, the kernel finally achieve 0.632 LB score  ... hope it is helpful!**\n\nHi everyone!\n\nFirst of all, thanks @daisukelab for his great kernel on preprocessed Mel-Spectrogram data. This data very much benefit all of participants here. However, currently only Pytorch-kernel implementation (such as one by @mhiro2 ) exists. \n\nIf anyone is like me who also would like to try this data based on Keras implementation, you can have a Keras starter kernel here : \n\nhttps://www.kaggle.com/ratthachat/fat19-kerasbaseline-on-preprocesseddata-lb576/\n\nHope it is helpful and happy kaggling everybody!",
      "votes": null
    },
    {
      "id": "526952",
      "postDate": "05/04/2019 07:23:40",
      "content": "<p>Thank you! \nI will compare and study your keras kernel with other pytorch kernels.</p>",
      "rawMarkdown": "Thank you! \nI will compare and study your keras kernel with other pytorch kernels.",
      "votes": null
    },
    {
      "id": "527259",
      "postDate": "05/05/2019 00:11:34",
      "content": "<p>Hi <a href=\"/dhaqui\">@dhaqui</a> , if there is any interesting results, please let me know!</p>",
      "rawMarkdown": "Hi @dhaqui , if there is any interesting results, please let me know!",
      "votes": null
    },
    {
      "id": "532169",
      "postDate": "05/16/2019 10:33:26",
      "content": "<p>Thanks for sharing! I'm into Keras implementation too.</p>",
      "rawMarkdown": "Thanks for sharing! I'm into Keras implementation too.",
      "votes": null
    },
    {
      "id": "546427",
      "postDate": "06/06/2019 15:24:29",
      "content": "<p>Thanks for the awesome kernel!\nIs tf_lwlrap metric supposed to go above 1.0?\nI copied tf_lwlrap to my model, and I got  1.06 for training metric and 0.8 for validation.</p>",
      "rawMarkdown": "Thanks for the awesome kernel!\nIs tf_lwlrap metric supposed to go above 1.0?\nI copied tf_lwlrap to my model, and I got  1.06 for training metric and 0.8 for validation.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 526952,
      "author_name": "dhaqui",
      "author_url": "",
      "post_date": "05/04/2019 07:23:40",
      "content": "<p>Thank you! \nI will compare and study your keras kernel with other pytorch kernels.</p>",
      "votes": null,
      "replies": [
        {
          "id": 527259,
          "author_name": "ratthachat",
          "author_url": "",
          "post_date": "05/05/2019 00:11:34",
          "content": "<p>Hi <a href=\"/dhaqui\">@dhaqui</a> , if there is any interesting results, please let me know!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 546427,
          "author_name": "chikim",
          "author_url": "",
          "post_date": "06/06/2019 15:24:29",
          "content": "<p>Thanks for the awesome kernel!\nIs tf_lwlrap metric supposed to go above 1.0?\nI copied tf_lwlrap to my model, and I got  1.06 for training metric and 0.8 for validation.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 532169,
      "author_name": "vivaroma",
      "author_url": "",
      "post_date": "05/16/2019 10:33:26",
      "content": "<p>Thanks for sharing! I'm into Keras implementation too.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "526897": "**EDIT: Now, with a handful of augmentations, the kernel finally achieve 0.632 LB score  ... hope it is helpful!**\n\nHi everyone!\n\nFirst of all, thanks @daisukelab for his great kernel on preprocessed Mel-Spectrogram data. This data very much benefit all of participants here. However, currently only Pytorch-kernel implementation (such as one by @mhiro2 ) exists. \n\nIf anyone is like me who also would like to try this data based on Keras implementation, you can have a Keras starter kernel here : \n\nhttps://www.kaggle.com/ratthachat/fat19-kerasbaseline-on-preprocesseddata-lb576/\n\nHope it is helpful and happy kaggling everybody!",
    "526952": "Thank you! \nI will compare and study your keras kernel with other pytorch kernels.",
    "527259": "Hi @dhaqui , if there is any interesting results, please let me know!",
    "532169": "Thanks for sharing! I'm into Keras implementation too.",
    "546427": "Thanks for the awesome kernel!\nIs tf_lwlrap metric supposed to go above 1.0?\nI copied tf_lwlrap to my model, and I got  1.06 for training metric and 0.8 for validation."
  },
  "source": "meta"
}