@maggiemd @ryanholbrook
\nPlease clarify whether the following licenses are allowed for external datasets, source code, and pre-trained models.
Sorry if this has already been answered in other threads, but it has not been answered in this thread and this one.
\n@maggiemd @ryanholbrook
\nPlease clarify whether the following licenses are allowed for external datasets, source code, and pre-trained models.
Sorry if this has already been answered in other threads, but it has not been answered in this thread and this one.
\nHi all,
\nThanks for your patience on this, and for bringing it to our attention. Benetech has no intention of disallowing the use of CHART-Info / ICDAR datasets for this competition. As such, we are currently working with our legal team to resolve this issue by making sure the rules and winner license statement allow the use of this data. Your valuable time and effort working with this data will not go to waste.
\nMy apologies for any confusion this may have caused.
", "rawMarkdown": "Hi all, \n\nThanks for your patience on this, and for bringing it to our attention. Benetech has no intention of disallowing the use of CHART-Info / ICDAR datasets for this competition. As such, we are currently working with our legal team to resolve this issue by making sure the rules and winner license statement allow the use of this data. Your valuable time and effort working with this data will not go to waste. \n\nMy apologies for any confusion this may have caused.", "votes": 13, "replies": [ { "id": 2269130, "postDate": "2023-05-22T08:31:59.923Z", "content": "@benjiaa Thanks for the follow-up !
\nHow about models with non-commercial licenses such as https://github.com/ultralytics/yolov5 ?
", "rawMarkdown": "@benjiaa Thanks for the follow-up !\n\nHow about models with non-commercial licenses such as https://github.com/ultralytics/yolov5 ?", "votes": 6, "replies": [ { "id": 2270788, "postDate": "2023-05-23T12:29:51.273Z", "content": "I'm not an expert on licenses or anything but afaiu it allows for commercial use https://github.com/ultralytics/yolov5/blob/master/LICENSE
", "rawMarkdown": "I'm not an expert on licenses or anything but afaiu it allows for commercial use https://github.com/ultralytics/yolov5/blob/master/LICENSE", "votes": 2 }, { "id": 2278064, "postDate": "2023-05-28T09:47:41.137Z", "content": "@benjiaa could you comment on whether using models like yolov5/yolov7 allowed? and is the dataset you mentioned allowed?
", "rawMarkdown": "@benjiaa could you comment on whether using models like yolov5/yolov7 allowed? and is the dataset you mentioned allowed?", "votes": 1 }, { "id": 2284827, "postDate": "2023-06-02T09:49:12.730Z", "content": "@maggiemd could you share some light on this? There is 11 days to go and soon we will need to wrap up this comp.
", "rawMarkdown": "@maggiemd could you share some light on this? There is 11 days to go and soon we will need to wrap up this comp.", "votes": 3 } ] } ] }, { "id": 2255165, "postDate": "2023-05-11T14:44:45.443Z", "content": "@maggiemd @ryanholbrook @benjiaa
\nLet me add a question.
What is the intent of this statement in Dataset Description?
\nAdditional Resources\nThis competition was inspired by the CHART-Info competition series. You may find resources (such as additional training data) from those competitions useful.\nThe dataset for the CHART-Info 2022 competition is licensed under CC BY-NC-SA 4.0, which is not available for commercial use.
\nAre you allowing us to use this dataset? Or can we not use it because it is CC BY-NC-SA 4.0?
Note: The above text was silently changed today by the organizers.
", "rawMarkdown": "Note: The above text was silently changed today by the organizers.", "votes": 2, "replies": [ { "id": 2256712, "postDate": "2023-05-12T17:05:57.527Z", "content": "Wow, this may indicate that the CHART-info dataset is not allowed in the competition. 🤔
", "rawMarkdown": "Wow, this may indicate that the CHART-info dataset is not allowed in the competition. 🤔", "votes": 1 } ] } ] }, { "id": 2248698, "postDate": "2023-05-07T05:49:53.923Z", "content": "I too wish for you to provide a clear answer.
\nIn addition to the above, as I have brought up in a thread I started, I would also like an answer as to whether or not I can use something for which the license is not clearly stated.
My understanding is the following:
\nMIT & Apache are allowed. GPL isn't allowed. I don't know about CC BY-NC-SA 4.0.
\nThe key principle is that the licence should allow the organizer to use the competition code commercially (without open sourcing all its code). GPL doesn't allow this, whereas MIT & Apache do.
", "rawMarkdown": "My understanding is the following:\n\nMIT & Apache are allowed. GPL isn't allowed. I don't know about CC BY-NC-SA 4.0.\n\nThe key principle is that the licence should allow the organizer to use the competition code commercially (without open sourcing all its code). GPL doesn't allow this, whereas MIT & Apache do.", "votes": 2 }, { "id": 2255944, "postDate": "2023-05-12T06:03:34.030Z", "content": "Since the winning license requirement is MIT, your model should in \"no event limit(s) commercial use of such code or model containing or depending on such code\" The reference to the other effort was simply as an example of something to use as you're getting up to speed on what we are looking to do here. We are not making any statements about the licensing requirements of anything outside of Kaggle.
", "rawMarkdown": "Since the winning license requirement is MIT, your model should in \"no event limit(s) commercial use of such code or model containing or depending on such code\" The reference to the other effort was simply as an example of something to use as you're getting up to speed on what we are looking to do here. We are not making any statements about the licensing requirements of anything outside of Kaggle. ", "replies": [ { "id": 2256038, "postDate": "2023-05-12T08:25:12.850Z", "content": "Am I correct in understanding that we are not allowed to use the ICDAR dataset for training after all, either?
\n(Am I correct in assuming that solutions using that dataset will eventually be excluded from the LB, even if they are submitted?)
Originally, the data tab was worded \"(such as additional training data)\" as penguin46 quoted.
\nHowever, I just checked and that notation has been removed.
\nWas this removed because it was discovered that it should not actually be used for training?
At least I believed the above statement by you guys and spent a lot of time on how to use the ICDAR data.
\nI may be at fault for not checking the license first, but I hope you realize that your statements still carry a great deal of weight.
\nAnd while it is fine to change your mind in response to a point of view, I hope that if you have made a change, you will not make a silent correction, but will make sure to make it known, including your intentions.
No, no, it should certainly be available. All the top teams should be using it, if kaggle wrote it like that and then banned it, it would be a riot haha.
", "rawMarkdown": "No, no, it should certainly be available. All the top teams should be using it, if kaggle wrote it like that and then banned it, it would be a riot haha.", "votes": 1 }, { "id": 2256573, "postDate": "2023-05-12T15:33:29.350Z", "content": "Thank you for your response.
\nBut the answer from kaggle staff is vague and I am very confused.
I think the following is at least a summary of the information obtained.
\nConsidering all of these together, I think that the model trained using the ICDAR dataset is affected by the non-commercial license and therefore does not meet the competition entry requirements (for commercial use).
\nI am not very familiar with the law, so am I interpreting this in a way that I am wrong and it is in fact safe to use?
And most of all, I'm really curious as to why the kaggle staff is secretly modifying the text that was previously listed on the Data tab. (If there is nothing problm, why did they modify it?)
", "rawMarkdown": "Thank you for your response.\nBut the answer from kaggle staff is vague and I am very confused.\n\nI think the following is at least a summary of the information obtained.\n1. the ICDAR dataset is released under a CC BY-NC-SA 4.0 license (non-commercial license)\n2. the models you submit must not preclude any commercial use (response from Maggie)\n\nConsidering all of these together, I think that the model trained using the ICDAR dataset is affected by the non-commercial license and therefore does not meet the competition entry requirements (for commercial use).\nI am not very familiar with the law, so am I interpreting this in a way that I am wrong and it is in fact safe to use?\n\nAnd most of all, I'm really curious as to why the kaggle staff is secretly modifying the text that was previously listed on the Data tab. (If there is nothing problm, why did they modify it?)", "votes": 2 }, { "id": 2256583, "postDate": "2023-05-12T15:43:29.327Z", "content": "To add a few thoughts of my own, I believe that if we leave this vague as a rule, we will end up with a mix of people who did and did not use this data set, and the competition will no longer be viable.
\nOf course, I have spent a lot of time on this, so I honestly cannot be convinced if they say it cannot be used now, but more than that, I hope that the staff will clarify it properly so that the final result will be acceptable to all.
", "rawMarkdown": "To add a few thoughts of my own, I believe that if we leave this vague as a rule, we will end up with a mix of people who did and did not use this data set, and the competition will no longer be viable.\n\nOf course, I have spent a lot of time on this, so I honestly cannot be convinced if they say it cannot be used now, but more than that, I hope that the staff will clarify it properly so that the final result will be acceptable to all.", "votes": 3 }, { "id": 2256601, "postDate": "2023-05-12T15:53:28.577Z", "content": "There is no doubt that unclear decisions lead to confusion.
\nThe dataset contains approximately 4000 images of fully labeled data. (Including unlabeled, about 20,000 images!)
\nConsidering that the non-synthetic data for the given training dataset is only 1000 images, this is a very valuable dataset.
@kashiwaba @ryotayoshinobu I've emailed the guys from ICDAR to see under which license is the dataset. I haven't found anywhere in the Infographics competition webpage about the dataset's license. How did you know it was under CC BY-NC-SA 4.0?
Out of curiosity, how can Kaggle verify if you have used the dataset or not? Suppose you trained a model with the dataset and your final solution includes only the model weights. Are competitors required to reproduce model training and achieve the same result?
\nI'll wait until I get a response back. It would be unfair for some competitors to use a dataset which does not meet the competition's requirements.
", "rawMarkdown": "@kashiwaba @ryotayoshinobu I've emailed the guys from ICDAR to see under which license is the dataset. I haven't found anywhere in the Infographics competition webpage about the dataset's license. How did you know it was under `CC BY-NC-SA 4.0`? \n\nOut of curiosity, how can Kaggle verify if you have used the dataset or not? Suppose you trained a model with the dataset and your final solution includes only the model weights. Are competitors required to reproduce model training and achieve the same result?\n\nI'll wait until I get a response back. It would be unfair for some competitors to use a dataset which does not meet the competition's requirements." }, { "id": 2256924, "postDate": "2023-05-12T20:51:52.113Z", "content": "\n\n\nHow did you know it was under CC BY-NC-SA 4.0?
\n
The license can be found when you download the dataset, which includes license.txt and clearly states that it is CC BY-NC-SA 4.0. (\"This work is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-sa/4.0/ or send a letter to Creative Commons, PO Box 1866, Mountain View, CA 94042, USA.\")
\n\n\nhow can Kaggle verify if you have used the dataset or not?
\n
The winners of the competition are required to submit source code that can generate an award-winning submission (11. WINNERS OBLIGATIONS (a)). I am not sure if the training code is included here. If only the inference code is needed, it will not be found out. As for the other participants, kaggle will not be able to know what data they trained on.
", "rawMarkdown": "@alejopaullier \n\n> How did you know it was under CC BY-NC-SA 4.0?\n\nThe license can be found when you download the dataset, which includes license.txt and clearly states that it is CC BY-NC-SA 4.0. (\"This work is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-sa/4.0/ or send a letter to Creative Commons, PO Box 1866, Mountain View, CA 94042, USA.\")\n\n> how can Kaggle verify if you have used the dataset or not?\n\nThe winners of the competition are required to submit source code that can generate an award-winning submission (11. WINNERS OBLIGATIONS (a)). I am not sure if the training code is included here. If only the inference code is needed, it will not be found out. As for the other participants, kaggle will not be able to know what data they trained on.", "votes": 1 } ] } ] }, { "id": 2289027, "postDate": "2023-06-05T21:30:10.133Z", "content": "Hi All,
\nPlease see our response in the newly inserted Section A. 3 as well as the new pinned discussion post on this matter that can be found here: https://www.kaggle.com/competitions/benetech-making-graphs-accessible/discussion/415275
Hi all,
\nThanks for your patience on this, and for bringing it to our attention. Benetech has no intention of disallowing the use of CHART-Info / ICDAR datasets for this competition. As such, we are currently working with our legal team to resolve this issue by making sure the rules and winner license statement allow the use of this data. Your valuable time and effort working with this data will not go to waste.
\nMy apologies for any confusion this may have caused.
", "votes": 13, "replies": [ { "id": 2269130, "author_name": "Theo Viel", "author_url": "", "post_date": "2023-05-22T08:31:59.923000", "content": "@benjiaa Thanks for the follow-up !
\nHow about models with non-commercial licenses such as https://github.com/ultralytics/yolov5 ?
", "votes": 6, "replies": [ { "id": 2270788, "author_name": "Eduardo Rocha de Andrade", "author_url": "", "post_date": "2023-05-23T12:29:51.273000", "content": "I'm not an expert on licenses or anything but afaiu it allows for commercial use https://github.com/ultralytics/yolov5/blob/master/LICENSE
", "votes": 2, "replies": [] }, { "id": 2278064, "author_name": "CroDoc", "author_url": "", "post_date": "2023-05-28T09:47:41.137000", "content": "@benjiaa could you comment on whether using models like yolov5/yolov7 allowed? and is the dataset you mentioned allowed?
", "votes": 1, "replies": [] }, { "id": 2284827, "author_name": "CroDoc", "author_url": "", "post_date": "2023-06-02T09:49:12.730000", "content": "@maggiemd could you share some light on this? There is 11 days to go and soon we will need to wrap up this comp.
", "votes": 3, "replies": [] } ] } ] }, { "id": 2255165, "author_name": "penguin46", "author_url": "", "post_date": "2023-05-11T14:44:45.443000", "content": "@maggiemd @ryanholbrook @benjiaa
\nLet me add a question.
What is the intent of this statement in Dataset Description?
\nAdditional Resources\nThis competition was inspired by the CHART-Info competition series. You may find resources (such as additional training data) from those competitions useful.\nThe dataset for the CHART-Info 2022 competition is licensed under CC BY-NC-SA 4.0, which is not available for commercial use.
\nAre you allowing us to use this dataset? Or can we not use it because it is CC BY-NC-SA 4.0?
Note: The above text was silently changed today by the organizers.
", "votes": 2, "replies": [ { "id": 2256712, "author_name": "june", "author_url": "", "post_date": "2023-05-12T17:05:57.527000", "content": "Wow, this may indicate that the CHART-info dataset is not allowed in the competition. 🤔
", "votes": 1, "replies": [] } ] } ] }, { "id": 2248698, "author_name": "YumeNeko", "author_url": "", "post_date": "2023-05-07T05:49:53.923000", "content": "I too wish for you to provide a clear answer.
\nIn addition to the above, as I have brought up in a thread I started, I would also like an answer as to whether or not I can use something for which the license is not clearly stated.
My understanding is the following:
\nMIT & Apache are allowed. GPL isn't allowed. I don't know about CC BY-NC-SA 4.0.
\nThe key principle is that the licence should allow the organizer to use the competition code commercially (without open sourcing all its code). GPL doesn't allow this, whereas MIT & Apache do.
", "votes": 2, "replies": [] }, { "id": 2255944, "author_name": "Maggie", "author_url": "", "post_date": "2023-05-12T06:03:34.030000", "content": "Since the winning license requirement is MIT, your model should in \"no event limit(s) commercial use of such code or model containing or depending on such code\" The reference to the other effort was simply as an example of something to use as you're getting up to speed on what we are looking to do here. We are not making any statements about the licensing requirements of anything outside of Kaggle.
", "votes": 0, "replies": [ { "id": 2256038, "author_name": "YumeNeko", "author_url": "", "post_date": "2023-05-12T08:25:12.850000", "content": "Am I correct in understanding that we are not allowed to use the ICDAR dataset for training after all, either?
\n(Am I correct in assuming that solutions using that dataset will eventually be excluded from the LB, even if they are submitted?)
Originally, the data tab was worded \"(such as additional training data)\" as penguin46 quoted.
\nHowever, I just checked and that notation has been removed.
\nWas this removed because it was discovered that it should not actually be used for training?
At least I believed the above statement by you guys and spent a lot of time on how to use the ICDAR data.
\nI may be at fault for not checking the license first, but I hope you realize that your statements still carry a great deal of weight.
\nAnd while it is fine to change your mind in response to a point of view, I hope that if you have made a change, you will not make a silent correction, but will make sure to make it known, including your intentions.
No, no, it should certainly be available. All the top teams should be using it, if kaggle wrote it like that and then banned it, it would be a riot haha.
", "votes": 1, "replies": [] }, { "id": 2256573, "author_name": "YumeNeko", "author_url": "", "post_date": "2023-05-12T15:33:29.350000", "content": "Thank you for your response.
\nBut the answer from kaggle staff is vague and I am very confused.
I think the following is at least a summary of the information obtained.
\nConsidering all of these together, I think that the model trained using the ICDAR dataset is affected by the non-commercial license and therefore does not meet the competition entry requirements (for commercial use).
\nI am not very familiar with the law, so am I interpreting this in a way that I am wrong and it is in fact safe to use?
And most of all, I'm really curious as to why the kaggle staff is secretly modifying the text that was previously listed on the Data tab. (If there is nothing problm, why did they modify it?)
", "votes": 2, "replies": [] }, { "id": 2256583, "author_name": "YumeNeko", "author_url": "", "post_date": "2023-05-12T15:43:29.327000", "content": "To add a few thoughts of my own, I believe that if we leave this vague as a rule, we will end up with a mix of people who did and did not use this data set, and the competition will no longer be viable.
\nOf course, I have spent a lot of time on this, so I honestly cannot be convinced if they say it cannot be used now, but more than that, I hope that the staff will clarify it properly so that the final result will be acceptable to all.
", "votes": 3, "replies": [] }, { "id": 2256601, "author_name": "penguin46", "author_url": "", "post_date": "2023-05-12T15:53:28.577000", "content": "There is no doubt that unclear decisions lead to confusion.
\nThe dataset contains approximately 4000 images of fully labeled data. (Including unlabeled, about 20,000 images!)
\nConsidering that the non-synthetic data for the given training dataset is only 1000 images, this is a very valuable dataset.
@kashiwaba @ryotayoshinobu I've emailed the guys from ICDAR to see under which license is the dataset. I haven't found anywhere in the Infographics competition webpage about the dataset's license. How did you know it was under CC BY-NC-SA 4.0?
Out of curiosity, how can Kaggle verify if you have used the dataset or not? Suppose you trained a model with the dataset and your final solution includes only the model weights. Are competitors required to reproduce model training and achieve the same result?
\nI'll wait until I get a response back. It would be unfair for some competitors to use a dataset which does not meet the competition's requirements.
", "votes": 0, "replies": [] }, { "id": 2256924, "author_name": "penguin46", "author_url": "", "post_date": "2023-05-12T20:51:52.113000", "content": "\n\n\nHow did you know it was under CC BY-NC-SA 4.0?
\n
The license can be found when you download the dataset, which includes license.txt and clearly states that it is CC BY-NC-SA 4.0. (\"This work is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-sa/4.0/ or send a letter to Creative Commons, PO Box 1866, Mountain View, CA 94042, USA.\")
\n\n\nhow can Kaggle verify if you have used the dataset or not?
\n
The winners of the competition are required to submit source code that can generate an award-winning submission (11. WINNERS OBLIGATIONS (a)). I am not sure if the training code is included here. If only the inference code is needed, it will not be found out. As for the other participants, kaggle will not be able to know what data they trained on.
", "votes": 1, "replies": [] } ] } ] }, { "id": 2289027, "author_name": "benjiaa", "author_url": "", "post_date": "2023-06-05T21:30:10.133000", "content": "Hi All,
\nPlease see our response in the newly inserted Section A. 3 as well as the new pinned discussion post on this matter that can be found here: https://www.kaggle.com/competitions/benetech-making-graphs-accessible/discussion/415275