{
  "id": 110045,
  "title": "objects365 dataset",
  "url": "/competitions/open-images-2019-object-detection/discussion/110045",
  "author_name": "",
  "post_date": "2019-09-24T10:49:36.558270Z",
  "votes": 26,
  "comment_count": 17,
  "views": 0,
  "content": "<p><a href=\"/juliaelliott\">@juliaelliott</a> , I have noticed in the external data thread that several participants have used objects365 dataset. According to the competition rules, this is not allowed as it is only available for academic research and not to the general public (i.e. you can't even request access to it if you only have gmail address). See 7.c in the rules (... you will (i) ensure the External Data is available for use by all participants...). </p>",
  "messages": [
    {
      "id": "633017",
      "postDate": "09/24/2019 10:49:36",
      "content": "<p><a href=\"/juliaelliott\">@juliaelliott</a> , I have noticed in the external data thread that several participants have used objects365 dataset. According to the competition rules, this is not allowed as it is only available for academic research and not to the general public (i.e. you can't even request access to it if you only have gmail address). See 7.c in the rules (... you will (i) ensure the External Data is available for use by all participants...). </p>",
      "rawMarkdown": "juliaelliott , I have noticed in the external data thread that several participants have used objects365 dataset. According to the competition rules, this is not allowed as it is only available for academic research and not to the general public (i.e. you can't even request access to it if you only have gmail address). See 7.c in the rules (... you will (i) ensure the External Data is available for use by all participants...).",
      "votes": null
    },
    {
      "id": "634758",
      "postDate": "09/26/2019 18:12:35",
      "content": "<p>As far as I can tell, the Objects365 dataset's annotations are licensed under CCBY 4.0 with attribution. Based on those terms, the dataset is available for all to use. The images are subject to the copyright terms of the original images (not owned by Objects365), so likewise available for all to use. I'm not seeing an issue here.</p>",
      "rawMarkdown": "As far as I can tell, the Objects365 dataset's annotations are licensed under CCBY 4.0 with attribution. Based on those terms, the dataset is available for all to use. The images are subject to the copyright terms of the original images (not owned by Objects365), so likewise available for all to use. I'm not seeing an issue here.",
      "votes": null
    },
    {
      "id": "634765",
      "postDate": "09/26/2019 18:21:03",
      "content": "<p>Try to get access to it with your gmail... It is only for people from \"domains\".</p>",
      "rawMarkdown": "Try to get access to it with your gmail... It is only for people from \"domains\".",
      "votes": null
    },
    {
      "id": "634768",
      "postDate": "09/26/2019 18:26:13",
      "content": "<p>Here is the registration blurb:</p>\n\n<p>To apply the Objects365 dataset, please agree on the license and provide the below information via email. We will only take applications from organization email (please DO NOT use the emails from gmail/163/qq). Anyone who uses the Objects365 dataset should obey the license and send us an email for registration.</p>",
      "rawMarkdown": "Here is the registration blurb:\n\nTo apply the Objects365 dataset, please agree on the license and provide the below information via email. We will only take applications from organization email (please DO NOT use the emails from gmail/163/qq). Anyone who uses the Objects365 dataset should obey the license and send us an email for registration.",
      "votes": null
    },
    {
      "id": "634971",
      "postDate": "09/27/2019 02:22:39",
      "content": "<p>Can the organizers look into this?</p>",
      "rawMarkdown": "Can the organizers look into this?",
      "votes": null
    },
    {
      "id": "635585",
      "postDate": "09/27/2019 21:53:57",
      "content": "<p>Our position is not to recommend the use of any particular dataset, but we are not aware of any reasons not to use it for purposes of the competition.</p>",
      "rawMarkdown": "Our position is not to recommend the use of any particular dataset, but we are not aware of any reasons not to use it for purposes of the competition.",
      "votes": null
    },
    {
      "id": "635592",
      "postDate": "09/27/2019 22:17:32",
      "content": "<p><a href=\"/juliaelliott\">@juliaelliott</a>, it's not about your position, it's about the rules. As this dataset is not accessible to me (having no organization email), the rules explicitly forbid it's use. Mund you, this is not a small thing. Getting high qualuty annotated images for the low count classes can mean few percent in this competition. </p>",
      "rawMarkdown": "juliaelliott, it's not about your position, it's about the rules. As this dataset is not accessible to me (having no organization email), the rules explicitly forbid it's use. Mund you, this is not a small thing. Getting high qualuty annotated images for the low count classes can mean few percent in this competition.",
      "votes": null
    },
    {
      "id": "636629",
      "postDate": "09/29/2019 23:28:40",
      "content": "<p><a href=\"/juliaelliott\">@juliaelliott</a>\n<a href=\"/akuznetsa\">@akuznetsa</a> </p>\n\n<p>Some people have access to the dataset. Some people don't.</p>\n\n<p>This probably violates the rules Moshel mentioned.</p>\n\n<p>Could you please answer to this potential violation of the rules? I agree that the license is not the problem here.</p>",
      "rawMarkdown": "juliaelliott\n@akuznetsa \n\nSome people have access to the dataset. Some people don't.\n\nThis probably violates the rules Moshel mentioned.\n\nCould you please answer to this potential violation of the rules? I agree that the license is not the problem here.",
      "votes": null
    },
    {
      "id": "637273",
      "postDate": "09/30/2019 20:54:50",
      "content": "<p><a href=\"/juliaelliott\">@juliaelliott</a> , <a href=\"/dansbecker\">@dansbecker</a> , we are left hanging here, with no responses from the team. Even \"we are looking into this\" is something... only a day to go</p>",
      "rawMarkdown": "juliaelliott , @dansbecker , we are left hanging here, with no responses from the team. Even \"we are looking into this\" is something... only a day to go",
      "votes": null
    },
    {
      "id": "637704",
      "postDate": "10/01/2019 07:58:26",
      "content": "<p>This does not seem to get lots of attention but this is a serious one as it probably largely affects competition's results.</p>\n\n<p>From <a href=\"https://www.kaggle.com/c/open-images-2019-object-detection/rules\">https://www.kaggle.com/c/open-images-2019-object-detection/rules</a></p>\n\n<blockquote>\n  <p>However, you will (i) ensure the External Data is available to use by all participants of the competition for purposes of the competition at no cost to the other participants</p>\n</blockquote>\n\n<p>Objects365 dataset is <strong>not</strong> available to use by all participants of the competition.</p>\n\n<p>Why is this ignored or is this discussed somewhere? </p>",
      "rawMarkdown": "This does not seem to get lots of attention but this is a serious one as it probably largely affects competition's results.\n\nFrom https://www.kaggle.com/c/open-images-2019-object-detection/rules\n&gt; However, you will (i) ensure the External Data is available to use by all participants of the competition for purposes of the competition at no cost to the other participants\n\nObjects365 dataset is **not** available to use by all participants of the competition.\n\nWhy is this ignored or is this discussed somewhere?",
      "votes": null
    },
    {
      "id": "637728",
      "postDate": "10/01/2019 08:13:30",
      "content": "<p>Dear all, thanks for raising the issue. We are looking into the situation with this dataset as it is a little unusual. Please be patient, we will make sure to provide the final decision here as soon as possible.</p>",
      "rawMarkdown": "Dear all, thanks for raising the issue. We are looking into the situation with this dataset as it is a little unusual. Please be patient, we will make sure to provide the final decision here as soon as possible.",
      "votes": null
    },
    {
      "id": "637869",
      "postDate": "10/01/2019 10:25:35",
      "content": "<p>We made the decision to allow the Object365 dataset to remain permitted for use as external data, based on the published CCBY license (which allows everyone to use it). We have come to realize that the dataset owner’s distribution of the dataset may not be in alignment with the stated dataset licensing and have contacted them to resolve this issue. However, the teams using the dataset were likely unaware of the issue and hence we will not induce any penalties or disqualifications based on the usage of Object365 dataset in Open Images Challenge 2019</p>",
      "rawMarkdown": "We made the decision to allow the Object365 dataset to remain permitted for use as external data, based on the published CCBY license (which allows everyone to use it). We have come to realize that the dataset owner’s distribution of the dataset may not be in alignment with the stated dataset licensing and have contacted them to resolve this issue. However, the teams using the dataset were likely unaware of the issue and hence we will not induce any penalties or disqualifications based on the usage of Object365 dataset in Open Images Challenge 2019",
      "votes": null
    },
    {
      "id": "637897",
      "postDate": "10/01/2019 10:56:02",
      "content": "<p>Participants have spent a lot of time and resources on this. Didn't expect such a reply from the organizers. Anyways, no issues. :) \nHopefully, such things doesnt happen in the future</p>",
      "rawMarkdown": "Participants have spent a lot of time and resources on this. Didn't expect such a reply from the organizers. Anyways, no issues. :) \nHopefully, such things doesnt happen in the future",
      "votes": null
    },
    {
      "id": "637910",
      "postDate": "10/01/2019 11:13:41",
      "content": "<p>Thank you for answering. \nI feel it's kind of strage that not knowing the rules can be a good reason to be excused from obeying the rules.</p>",
      "rawMarkdown": "Thank you for answering. \nI feel it's kind of strage that not knowing the rules can be a good reason to be excused from obeying the rules.",
      "votes": null
    },
    {
      "id": "637924",
      "postDate": "10/01/2019 11:21:57",
      "content": "<p>Thanks for addressing this. Although I find it hard to believe that you could ignore the blurb, i guess benefit of doubt prevail.</p>\n\n<p>It would be nice if kaggle will have proper procedures in place for monitoring requests from competition participants in the forum and notifying them that someone is looking into it. This was very awkward handling. </p>",
      "rawMarkdown": "Thanks for addressing this. Although I find it hard to believe that you could ignore the blurb, i guess benefit of doubt prevail.\n\nIt would be nice if kaggle will have proper procedures in place for monitoring requests from competition participants in the forum and notifying them that someone is looking into it. This was very awkward handling.",
      "votes": null
    },
    {
      "id": "638358",
      "postDate": "10/01/2019 20:01:51",
      "content": "<p>Plausible deniability... </p>",
      "rawMarkdown": "Plausible deniability...",
      "votes": null
    },
    {
      "id": "638509",
      "postDate": "10/02/2019 01:28:46",
      "content": "<p>I can't download it. NOT FAIR.</p>",
      "rawMarkdown": "I can't download it. NOT FAIR.",
      "votes": null
    },
    {
      "id": "638765",
      "postDate": "10/02/2019 11:54:06",
      "content": "<p>I agree, totally unfair ...</p>",
      "rawMarkdown": "I agree, totally unfair ...",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 634758,
      "author_name": "juliaelliott",
      "author_url": "",
      "post_date": "09/26/2019 18:12:35",
      "content": "<p>As far as I can tell, the Objects365 dataset's annotations are licensed under CCBY 4.0 with attribution. Based on those terms, the dataset is available for all to use. The images are subject to the copyright terms of the original images (not owned by Objects365), so likewise available for all to use. I'm not seeing an issue here.</p>",
      "votes": null,
      "replies": [
        {
          "id": 634765,
          "author_name": "moshel",
          "author_url": "",
          "post_date": "09/26/2019 18:21:03",
          "content": "<p>Try to get access to it with your gmail... It is only for people from \"domains\".</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 634768,
          "author_name": "moshel",
          "author_url": "",
          "post_date": "09/26/2019 18:26:13",
          "content": "<p>Here is the registration blurb:</p>\n\n<p>To apply the Objects365 dataset, please agree on the license and provide the below information via email. We will only take applications from organization email (please DO NOT use the emails from gmail/163/qq). Anyone who uses the Objects365 dataset should obey the license and send us an email for registration.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 636629,
          "author_name": "appian",
          "author_url": "",
          "post_date": "09/29/2019 23:28:40",
          "content": "<p><a href=\"/juliaelliott\">@juliaelliott</a>\n<a href=\"/akuznetsa\">@akuznetsa</a> </p>\n\n<p>Some people have access to the dataset. Some people don't.</p>\n\n<p>This probably violates the rules Moshel mentioned.</p>\n\n<p>Could you please answer to this potential violation of the rules? I agree that the license is not the problem here.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 634971,
      "author_name": "chandanverma",
      "author_url": "",
      "post_date": "09/27/2019 02:22:39",
      "content": "<p>Can the organizers look into this?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 635585,
      "author_name": "juliaelliott",
      "author_url": "",
      "post_date": "09/27/2019 21:53:57",
      "content": "<p>Our position is not to recommend the use of any particular dataset, but we are not aware of any reasons not to use it for purposes of the competition.</p>",
      "votes": null,
      "replies": [
        {
          "id": 635592,
          "author_name": "moshel",
          "author_url": "",
          "post_date": "09/27/2019 22:17:32",
          "content": "<p><a href=\"/juliaelliott\">@juliaelliott</a>, it's not about your position, it's about the rules. As this dataset is not accessible to me (having no organization email), the rules explicitly forbid it's use. Mund you, this is not a small thing. Getting high qualuty annotated images for the low count classes can mean few percent in this competition. </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 637273,
      "author_name": "moshel",
      "author_url": "",
      "post_date": "09/30/2019 20:54:50",
      "content": "<p><a href=\"/juliaelliott\">@juliaelliott</a> , <a href=\"/dansbecker\">@dansbecker</a> , we are left hanging here, with no responses from the team. Even \"we are looking into this\" is something... only a day to go</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 637704,
      "author_name": "appian",
      "author_url": "",
      "post_date": "10/01/2019 07:58:26",
      "content": "<p>This does not seem to get lots of attention but this is a serious one as it probably largely affects competition's results.</p>\n\n<p>From <a href=\"https://www.kaggle.com/c/open-images-2019-object-detection/rules\">https://www.kaggle.com/c/open-images-2019-object-detection/rules</a></p>\n\n<blockquote>\n  <p>However, you will (i) ensure the External Data is available to use by all participants of the competition for purposes of the competition at no cost to the other participants</p>\n</blockquote>\n\n<p>Objects365 dataset is <strong>not</strong> available to use by all participants of the competition.</p>\n\n<p>Why is this ignored or is this discussed somewhere? </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 637728,
      "author_name": "akuznetsa",
      "author_url": "",
      "post_date": "10/01/2019 08:13:30",
      "content": "<p>Dear all, thanks for raising the issue. We are looking into the situation with this dataset as it is a little unusual. Please be patient, we will make sure to provide the final decision here as soon as possible.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 637869,
      "author_name": "akuznetsa",
      "author_url": "",
      "post_date": "10/01/2019 10:25:35",
      "content": "<p>We made the decision to allow the Object365 dataset to remain permitted for use as external data, based on the published CCBY license (which allows everyone to use it). We have come to realize that the dataset owner’s distribution of the dataset may not be in alignment with the stated dataset licensing and have contacted them to resolve this issue. However, the teams using the dataset were likely unaware of the issue and hence we will not induce any penalties or disqualifications based on the usage of Object365 dataset in Open Images Challenge 2019</p>",
      "votes": null,
      "replies": [
        {
          "id": 637897,
          "author_name": "chandanverma",
          "author_url": "",
          "post_date": "10/01/2019 10:56:02",
          "content": "<p>Participants have spent a lot of time and resources on this. Didn't expect such a reply from the organizers. Anyways, no issues. :) \nHopefully, such things doesnt happen in the future</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 637910,
          "author_name": "appian",
          "author_url": "",
          "post_date": "10/01/2019 11:13:41",
          "content": "<p>Thank you for answering. \nI feel it's kind of strage that not knowing the rules can be a good reason to be excused from obeying the rules.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 637924,
          "author_name": "moshel",
          "author_url": "",
          "post_date": "10/01/2019 11:21:57",
          "content": "<p>Thanks for addressing this. Although I find it hard to believe that you could ignore the blurb, i guess benefit of doubt prevail.</p>\n\n<p>It would be nice if kaggle will have proper procedures in place for monitoring requests from competition participants in the forum and notifying them that someone is looking into it. This was very awkward handling. </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 638358,
          "author_name": "moshel",
          "author_url": "",
          "post_date": "10/01/2019 20:01:51",
          "content": "<p>Plausible deniability... </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 638509,
          "author_name": "artyomp",
          "author_url": "",
          "post_date": "10/02/2019 01:28:46",
          "content": "<p>I can't download it. NOT FAIR.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 638765,
          "author_name": "nhlr21",
          "author_url": "",
          "post_date": "10/02/2019 11:54:06",
          "content": "<p>I agree, totally unfair ...</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "633017": "juliaelliott , I have noticed in the external data thread that several participants have used objects365 dataset. According to the competition rules, this is not allowed as it is only available for academic research and not to the general public (i.e. you can't even request access to it if you only have gmail address). See 7.c in the rules (... you will (i) ensure the External Data is available for use by all participants...).",
    "634758": "As far as I can tell, the Objects365 dataset's annotations are licensed under CCBY 4.0 with attribution. Based on those terms, the dataset is available for all to use. The images are subject to the copyright terms of the original images (not owned by Objects365), so likewise available for all to use. I'm not seeing an issue here.",
    "634765": "Try to get access to it with your gmail... It is only for people from \"domains\".",
    "634768": "Here is the registration blurb:\n\nTo apply the Objects365 dataset, please agree on the license and provide the below information via email. We will only take applications from organization email (please DO NOT use the emails from gmail/163/qq). Anyone who uses the Objects365 dataset should obey the license and send us an email for registration.",
    "634971": "Can the organizers look into this?",
    "635585": "Our position is not to recommend the use of any particular dataset, but we are not aware of any reasons not to use it for purposes of the competition.",
    "635592": "juliaelliott, it's not about your position, it's about the rules. As this dataset is not accessible to me (having no organization email), the rules explicitly forbid it's use. Mund you, this is not a small thing. Getting high qualuty annotated images for the low count classes can mean few percent in this competition.",
    "636629": "juliaelliott\n@akuznetsa \n\nSome people have access to the dataset. Some people don't.\n\nThis probably violates the rules Moshel mentioned.\n\nCould you please answer to this potential violation of the rules? I agree that the license is not the problem here.",
    "637273": "juliaelliott , @dansbecker , we are left hanging here, with no responses from the team. Even \"we are looking into this\" is something... only a day to go",
    "637704": "This does not seem to get lots of attention but this is a serious one as it probably largely affects competition's results.\n\nFrom https://www.kaggle.com/c/open-images-2019-object-detection/rules\n&gt; However, you will (i) ensure the External Data is available to use by all participants of the competition for purposes of the competition at no cost to the other participants\n\nObjects365 dataset is **not** available to use by all participants of the competition.\n\nWhy is this ignored or is this discussed somewhere?",
    "637728": "Dear all, thanks for raising the issue. We are looking into the situation with this dataset as it is a little unusual. Please be patient, we will make sure to provide the final decision here as soon as possible.",
    "637869": "We made the decision to allow the Object365 dataset to remain permitted for use as external data, based on the published CCBY license (which allows everyone to use it). We have come to realize that the dataset owner’s distribution of the dataset may not be in alignment with the stated dataset licensing and have contacted them to resolve this issue. However, the teams using the dataset were likely unaware of the issue and hence we will not induce any penalties or disqualifications based on the usage of Object365 dataset in Open Images Challenge 2019",
    "637897": "Participants have spent a lot of time and resources on this. Didn't expect such a reply from the organizers. Anyways, no issues. :) \nHopefully, such things doesnt happen in the future",
    "637910": "Thank you for answering. \nI feel it's kind of strage that not knowing the rules can be a good reason to be excused from obeying the rules.",
    "637924": "Thanks for addressing this. Although I find it hard to believe that you could ignore the blurb, i guess benefit of doubt prevail.\n\nIt would be nice if kaggle will have proper procedures in place for monitoring requests from competition participants in the forum and notifying them that someone is looking into it. This was very awkward handling.",
    "638358": "Plausible deniability...",
    "638509": "I can't download it. NOT FAIR.",
    "638765": "I agree, totally unfair ..."
  },
  "source": "meta"
}