{
  "id": 108617,
  "title": "new dataset from chars in training set [link to dataset in post]",
  "url": "/competitions/kuzushiji-recognition/discussion/108617",
  "author_name": "",
  "post_date": "2019-09-12T19:45:13.821422900Z",
  "votes": 2,
  "comment_count": 7,
  "views": 0,
  "content": "<p>Hey <a href=\"/anokas\">@anokas</a>, <a href=\"/tkasasagi\">@tkasasagi</a>,\nthanks for the Kuzushiji Competition.\nI would like to create a dataset containing crops from all chars inside of the training set. However, the competition rules state the following: \"You agree not to transmit, duplicate, publish, redistribute or otherwise provide or make available the Competition Data to any party not participating in the Competition.\" Is there a way, that I create this dataset but not violate the competition rules?</p>\n\n<p>Kind regards,\nChris</p>\n\n<p>[update:] <a href=\"https://www.kaggle.com/christianwallenwein/kuzushiji-characters\">the dataset is now available here</a></p>",
  "messages": [
    {
      "id": "625211",
      "postDate": "09/12/2019 19:45:13",
      "content": "<p>Hey <a href=\"/anokas\">@anokas</a>, <a href=\"/tkasasagi\">@tkasasagi</a>,\nthanks for the Kuzushiji Competition.\nI would like to create a dataset containing crops from all chars inside of the training set. However, the competition rules state the following: \"You agree not to transmit, duplicate, publish, redistribute or otherwise provide or make available the Competition Data to any party not participating in the Competition.\" Is there a way, that I create this dataset but not violate the competition rules?</p>\n\n<p>Kind regards,\nChris</p>\n\n<p>[update:] <a href=\"https://www.kaggle.com/christianwallenwein/kuzushiji-characters\">the dataset is now available here</a></p>",
      "rawMarkdown": "Hey @anokas, @tkasasagi,\nthanks for the Kuzushiji Competition.\nI would like to create a dataset containing crops from all chars inside of the training set. However, the competition rules state the following: \"You agree not to transmit, duplicate, publish, redistribute or otherwise provide or make available the Competition Data to any party not participating in the Competition.\" Is there a way, that I create this dataset but not violate the competition rules?\n\nKind regards,\nChris\n\n[update:] [the dataset is now available here](https://www.kaggle.com/christianwallenwein/kuzushiji-characters)",
      "votes": null
    },
    {
      "id": "625661",
      "postDate": "09/13/2019 09:26:58",
      "content": "<p>Thank you for the question. For the dataset, we already have the cropped characters data for download online at CODH website. We just don't provide it in the competition because it's not necessary.</p>",
      "rawMarkdown": "Thank you for the question. For the dataset, we already have the cropped characters data for download online at CODH website. We just don't provide it in the competition because it's not necessary.",
      "votes": null
    },
    {
      "id": "626426",
      "postDate": "09/14/2019 09:50:40",
      "content": "<p>okay, thanks :)</p>",
      "rawMarkdown": "okay, thanks :)",
      "votes": null
    },
    {
      "id": "626503",
      "postDate": "09/14/2019 12:38:05",
      "content": "<p>Great Idea!!</p>",
      "rawMarkdown": "Great Idea!!",
      "votes": null
    },
    {
      "id": "627177",
      "postDate": "09/15/2019 14:34:02",
      "content": "<p>We would be fine with creation of such a dataset if the purpose is just to share within Kaggle forums/kernels etc.</p>",
      "rawMarkdown": "We would be fine with creation of such a dataset if the purpose is just to share within Kaggle forums/kernels etc.",
      "votes": null
    },
    {
      "id": "627226",
      "postDate": "09/15/2019 15:41:24",
      "content": "<p>okay, thanks</p>",
      "rawMarkdown": "okay, thanks",
      "votes": null
    },
    {
      "id": "631664",
      "postDate": "09/22/2019 12:43:15",
      "content": "<p>I have created a <a href=\"https://www.kaggle.com/frlemarchand/synthetic-kmnist-pages\">new dataset of synthetic Kuzushiji-filled pages</a> based the KMNIST dataset. As far as I know, it doesn't breach any licensing rights as I purely used the external dataset and used the same license. <a href=\"/anokas\">@anokas</a> <a href=\"/tkasasagi\">@tkasasagi</a> Could you please confirm it is all good (all details are in <a href=\"https://www.kaggle.com/frlemarchand/kuzushiji-page-generator\">this kernel</a>)? :)\nI haven't had time to use it yet, but I believe it could improve performance on character detection like in <a href=\"https://www.kaggle.com/hocop1/unet-character-detector\">this kernel</a>.</p>",
      "rawMarkdown": "I have created a [new dataset of synthetic Kuzushiji-filled pages](https://www.kaggle.com/frlemarchand/synthetic-kmnist-pages) based the KMNIST dataset. As far as I know, it doesn't breach any licensing rights as I purely used the external dataset and used the same license. @anokas @tkasasagi Could you please confirm it is all good (all details are in [this kernel](https://www.kaggle.com/frlemarchand/kuzushiji-page-generator))? :)\nI haven't had time to use it yet, but I believe it could improve performance on character detection like in [this kernel](https://www.kaggle.com/hocop1/unet-character-detector).",
      "votes": null
    },
    {
      "id": "631804",
      "postDate": "09/22/2019 17:54:15",
      "content": "<p><a href=\"/lodetomasi1995\">@lodetomasi1995</a> Hey, I just created the dataset. It's available here: <a href=\"https://www.kaggle.com/christianwallenwein/kuzushiji-characters\">https://www.kaggle.com/christianwallenwein/kuzushiji-characters</a></p>",
      "rawMarkdown": "lodetomasi1995 Hey, I just created the dataset. It's available here: [https://www.kaggle.com/christianwallenwein/kuzushiji-characters](https://www.kaggle.com/christianwallenwein/kuzushiji-characters)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 625661,
      "author_name": "tkasasagi",
      "author_url": "",
      "post_date": "09/13/2019 09:26:58",
      "content": "<p>Thank you for the question. For the dataset, we already have the cropped characters data for download online at CODH website. We just don't provide it in the competition because it's not necessary.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 626426,
      "author_name": "christianwallenwein",
      "author_url": "",
      "post_date": "09/14/2019 09:50:40",
      "content": "<p>okay, thanks :)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 626503,
      "author_name": "lodetomasi1995",
      "author_url": "",
      "post_date": "09/14/2019 12:38:05",
      "content": "<p>Great Idea!!</p>",
      "votes": null,
      "replies": [
        {
          "id": 631804,
          "author_name": "christianwallenwein",
          "author_url": "",
          "post_date": "09/22/2019 17:54:15",
          "content": "<p><a href=\"/lodetomasi1995\">@lodetomasi1995</a> Hey, I just created the dataset. It's available here: <a href=\"https://www.kaggle.com/christianwallenwein/kuzushiji-characters\">https://www.kaggle.com/christianwallenwein/kuzushiji-characters</a></p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 627177,
      "author_name": "anokas",
      "author_url": "",
      "post_date": "09/15/2019 14:34:02",
      "content": "<p>We would be fine with creation of such a dataset if the purpose is just to share within Kaggle forums/kernels etc.</p>",
      "votes": null,
      "replies": [
        {
          "id": 627226,
          "author_name": "christianwallenwein",
          "author_url": "",
          "post_date": "09/15/2019 15:41:24",
          "content": "<p>okay, thanks</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 631664,
      "author_name": "frlemarchand",
      "author_url": "",
      "post_date": "09/22/2019 12:43:15",
      "content": "<p>I have created a <a href=\"https://www.kaggle.com/frlemarchand/synthetic-kmnist-pages\">new dataset of synthetic Kuzushiji-filled pages</a> based the KMNIST dataset. As far as I know, it doesn't breach any licensing rights as I purely used the external dataset and used the same license. <a href=\"/anokas\">@anokas</a> <a href=\"/tkasasagi\">@tkasasagi</a> Could you please confirm it is all good (all details are in <a href=\"https://www.kaggle.com/frlemarchand/kuzushiji-page-generator\">this kernel</a>)? :)\nI haven't had time to use it yet, but I believe it could improve performance on character detection like in <a href=\"https://www.kaggle.com/hocop1/unet-character-detector\">this kernel</a>.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "625211": "Hey @anokas, @tkasasagi,\nthanks for the Kuzushiji Competition.\nI would like to create a dataset containing crops from all chars inside of the training set. However, the competition rules state the following: \"You agree not to transmit, duplicate, publish, redistribute or otherwise provide or make available the Competition Data to any party not participating in the Competition.\" Is there a way, that I create this dataset but not violate the competition rules?\n\nKind regards,\nChris\n\n[update:] [the dataset is now available here](https://www.kaggle.com/christianwallenwein/kuzushiji-characters)",
    "625661": "Thank you for the question. For the dataset, we already have the cropped characters data for download online at CODH website. We just don't provide it in the competition because it's not necessary.",
    "626426": "okay, thanks :)",
    "626503": "Great Idea!!",
    "627177": "We would be fine with creation of such a dataset if the purpose is just to share within Kaggle forums/kernels etc.",
    "627226": "okay, thanks",
    "631664": "I have created a [new dataset of synthetic Kuzushiji-filled pages](https://www.kaggle.com/frlemarchand/synthetic-kmnist-pages) based the KMNIST dataset. As far as I know, it doesn't breach any licensing rights as I purely used the external dataset and used the same license. @anokas @tkasasagi Could you please confirm it is all good (all details are in [this kernel](https://www.kaggle.com/frlemarchand/kuzushiji-page-generator))? :)\nI haven't had time to use it yet, but I believe it could improve performance on character detection like in [this kernel](https://www.kaggle.com/hocop1/unet-character-detector).",
    "631804": "lodetomasi1995 Hey, I just created the dataset. It's available here: [https://www.kaggle.com/christianwallenwein/kuzushiji-characters](https://www.kaggle.com/christianwallenwein/kuzushiji-characters)"
  },
  "source": "meta"
}