{ "id": 490622, "title": "Question about license", "url": "/competitions/image-matching-challenge-2024/discussion/490622", "author_name": "", "post_date": "2024-04-03T01:21:45.431497900Z", "votes": 13, "comment_count": 4, "views": 0, "content": "

After reviewing the rules, I noticed that IMC2024 has some differences compared to IMC2022 and IMC2023. It appears that the Winner License this year must be Apache 2.0. Therefore, I have a few questions regarding the license.

\n
    \n

  1. is the MIT license allowed?
    \nMy understanding is that Apache 2.0 and MIT are compatible licenses, meaning that software published under the MIT license can also be used. Is my understanding correct?

    2. \n

  2. (If the answer to 1. is Yes) Can DKM be used in this competition?
    \nI checked the license for DKM, which was frequently used in previous competitions, and it states that only models trained with Megadepth will be released under the MIT license.
    \nIn the experiments folder of the official repository, it seems that DKMv3_outdoor, which is also used in this Notebook, was trained with Megadepth. So, I am assuming it can be used, is that correct?

    3. \n

  3. am I correct in assuming that SuperGlue cannot be used?
    \nUnfortunately, SuperGlue, which was popular in previous competitions, is released under a license that does not allow commercial use. Therefore, I understand that SuperGlue cannot be used in this competition, at least for the official repositories. Is my assumption correct?

    4. \n
\n

I would be very grateful if you could answer these questions.

", "messages": [ { "id": "2732056", "postDate": "04/03/2024 01:21:45", "content": "

After reviewing the rules, I noticed that IMC2024 has some differences compared to IMC2022 and IMC2023. It appears that the Winner License this year must be Apache 2.0. Therefore, I have a few questions regarding the license.

\n
    \n

  1. is the MIT license allowed?
    \nMy understanding is that Apache 2.0 and MIT are compatible licenses, meaning that software published under the MIT license can also be used. Is my understanding correct?

    2. \n

  2. (If the answer to 1. is Yes) Can DKM be used in this competition?
    \nI checked the license for DKM, which was frequently used in previous competitions, and it states that only models trained with Megadepth will be released under the MIT license.
    \nIn the experiments folder of the official repository, it seems that DKMv3_outdoor, which is also used in this Notebook, was trained with Megadepth. So, I am assuming it can be used, is that correct?

    3. \n

  3. am I correct in assuming that SuperGlue cannot be used?
    \nUnfortunately, SuperGlue, which was popular in previous competitions, is released under a license that does not allow commercial use. Therefore, I understand that SuperGlue cannot be used in this competition, at least for the official repositories. Is my assumption correct?

    4. \n
\n

I would be very grateful if you could answer these questions.

", "rawMarkdown": "After reviewing the rules, I noticed that IMC2024 has some differences compared to IMC2022 and IMC2023. It appears that the Winner License this year must be Apache 2.0. Therefore, I have a few questions regarding the license.\n\n1. is the MIT license allowed?\nMy understanding is that Apache 2.0 and MIT are compatible licenses, meaning that software published under the MIT license can also be used. Is my understanding correct?\n\n2. (If the answer to 1. is Yes) Can DKM be used in this competition?\nI checked [the license for DKM](https://github.com/Parskatt/DKM/blob/main/LICENSE), which was frequently used in previous competitions, and it states that only models trained with Megadepth will be released under the MIT license. \nIn the experiments folder of [the official repository](https://github.com/Parskatt/DKM), it seems that DKMv3_outdoor, which is also used in [this Notebook](https://www.kaggle.com/code/motono0223/imagematching2024-dkm-example), was trained with Megadepth. So, I am assuming it can be used, is that correct?\n\n3. am I correct in assuming that SuperGlue cannot be used?\nUnfortunately, [SuperGlue](https://github.com/magicleap/SuperGluePretrainedNetwork), which was popular in previous competitions, is released under a license that does not allow commercial use. Therefore, I understand that SuperGlue cannot be used in this competition, at least for the official repositories. Is my assumption correct?\n\nI would be very grateful if you could answer these questions.", "votes": null }, { "id": "2732648", "postDate": "04/03/2024 08:51:26", "content": "

Hi,

\n

I believe, that your questions has been already answered in the Welcome thread: https://www.kaggle.com/competitions/image-matching-challenge-2024/discussion/486584

\n
\n

Prize-eligibility
\n Any submission, using non-commercial licensed 3rd party code is not prize eligible. E.g. submissions using SuperGlue (or SuperPoint) are not prize-eligible. Use LightGlue instead, which is both open source, and provides better performance.
\n GPL licensed 3rd party code is OK. However, your own code should be Apache 2.0 licensed.

\n
\n

However, one cannot have too much clarity, so I am happy to answer those again.
\nOne of the goals of this competition, is to promote and improve methods, which could be used by anyone freely. If you use this as a guideline principle, the answers are easy to predict:

\n
\n

is the MIT license allowed?
\n My understanding is that Apache 2.0 and MIT are compatible licenses, meaning that software published under the MIT license can also be used. Is my understanding correct?

\n
\n

Yes, you can use any MIT, or other OSS licensed software freely. Even GPLv3, which is not fully compatible with Apache 2.0 is OK. As long, as license allows commercial usage and open source, the 3rd party license is fine.
\nIt is your code (for submission and training), which should be Apache 2.0.

\n
\n

(If the answer to 1. is Yes) Can DKM be used in this competition?
\n I checked the license for DKM, which was frequently used in previous competitions, and it states that only models trained with Megadepth will be released under the MIT license.
\n In the experiments folder of the official repository, it seems that DKMv3_outdoor, which is also used in this Notebook, was trained with Megadepth. So, I am assuming it can be used, is that correct?

\n
\n

That is correct.

\n
\n

am I correct in assuming that SuperGlue cannot be used?
\n Unfortunately, SuperGlue, which was popular in previous competitions, is released under a license that does not allow commercial use. Therefore, I understand that SuperGlue cannot be used in this competition, at least for the official repositories. Is my assumption correct?

\n
\n

That is correct. You also cannot use official SuperPoint, licensed same as SuperGlue.
\nHowever, you can use open sourced SuperPoint https://github.com/rpautrat/SuperPoint, for example.
\nYou might be also interested in this topic https://www.kaggle.com/competitions/image-matching-challenge-2024/discussion/490723

\n

--
\nBest, Dmytro.

", "rawMarkdown": "Hi,\n\nI believe, that your questions has been already answered in the Welcome thread: https://www.kaggle.com/competitions/image-matching-challenge-2024/discussion/486584\n\n\n>Prize-eligibility\nAny submission, using non-commercial licensed 3rd party code is not prize eligible. E.g. submissions using SuperGlue (or SuperPoint) are not prize-eligible. Use LightGlue instead, which is both open source, and provides better performance.\nGPL licensed 3rd party code is OK. However, your own code should be Apache 2.0 licensed.\n\n\nHowever, one cannot have too much clarity, so I am happy to answer those again. \nOne of the goals of this competition, is to promote and improve methods, which could be used by anyone freely. If you use this as a guideline principle, the answers are easy to predict:\n\n >is the MIT license allowed?\nMy understanding is that Apache 2.0 and MIT are compatible licenses, meaning that software published under the MIT license can also be used. Is my understanding correct?\n\nYes, you can use any MIT, or other OSS licensed software freely. Even GPLv3, which is not fully compatible with Apache 2.0 is OK. As long, as license allows commercial usage and open source, the 3rd party license is fine. \nIt is your code (for submission and training), which should be Apache 2.0. \n\n>(If the answer to 1. is Yes) Can DKM be used in this competition?\nI checked the license for DKM, which was frequently used in previous competitions, and it states that only models trained with Megadepth will be released under the MIT license.\nIn the experiments folder of the official repository, it seems that DKMv3_outdoor, which is also used in this Notebook, was trained with Megadepth. So, I am assuming it can be used, is that correct?\n\nThat is correct. \n\n>am I correct in assuming that SuperGlue cannot be used?\nUnfortunately, SuperGlue, which was popular in previous competitions, is released under a license that does not allow commercial use. Therefore, I understand that SuperGlue cannot be used in this competition, at least for the official repositories. Is my assumption correct?\n\nThat is correct. You also cannot use official SuperPoint, licensed same as SuperGlue. \nHowever, you can use open sourced SuperPoint https://github.com/rpautrat/SuperPoint, for example.\nYou might be also interested in this topic https://www.kaggle.com/competitions/image-matching-challenge-2024/discussion/490723\n\n\n--\nBest, Dmytro.", "votes": null }, { "id": "2732858", "postDate": "04/03/2024 11:18:02", "content": "

I missed that thread, my apologies.
\nThanks for the answer, very clarifying.

", "rawMarkdown": "I missed that thread, my apologies.\nThanks for the answer, very clarifying.", "votes": null }, { "id": "2732863", "postDate": "04/03/2024 11:26:08", "content": "

You are welcome. It is nice to see previous challenges winners to participate again :)

", "rawMarkdown": "You are welcome. It is nice to see previous challenges winners to participate again :)", "votes": null }, { "id": "2790336", "postDate": "05/03/2024 06:24:29", "content": "

Hi, @oldufo -san!

\n

Is it possible to directly use a model trained on ImageNet that is not for commercial use?
\nFor example, using the timm library to use a model trained on ImageNet and incorporating the results of a 1000-class scene classification into an algorithm, is this allowed?

\n

Using ImageNet for pre-training is often a topic of discussion, but I would like your opinion on the case of direct use.

\n

BR

", "rawMarkdown": "Hi, @oldufo -san!\n\nIs it possible to directly use a model trained on ImageNet that is not for commercial use? \nFor example, using the timm library to use a model trained on ImageNet and incorporating the results of a 1000-class scene classification into an algorithm, is this allowed?\n\nUsing ImageNet for pre-training is often a topic of discussion, but I would like your opinion on the case of direct use.\n\nBR", "votes": null } ], "comments": [ { "id": 2732648, "author_name": "oldufo", "author_url": "", "post_date": "04/03/2024 08:51:26", "content": "

Hi,

\n

I believe, that your questions has been already answered in the Welcome thread: https://www.kaggle.com/competitions/image-matching-challenge-2024/discussion/486584

\n
\n

Prize-eligibility
\n Any submission, using non-commercial licensed 3rd party code is not prize eligible. E.g. submissions using SuperGlue (or SuperPoint) are not prize-eligible. Use LightGlue instead, which is both open source, and provides better performance.
\n GPL licensed 3rd party code is OK. However, your own code should be Apache 2.0 licensed.

\n
\n

However, one cannot have too much clarity, so I am happy to answer those again.
\nOne of the goals of this competition, is to promote and improve methods, which could be used by anyone freely. If you use this as a guideline principle, the answers are easy to predict:

\n
\n

is the MIT license allowed?
\n My understanding is that Apache 2.0 and MIT are compatible licenses, meaning that software published under the MIT license can also be used. Is my understanding correct?

\n
\n

Yes, you can use any MIT, or other OSS licensed software freely. Even GPLv3, which is not fully compatible with Apache 2.0 is OK. As long, as license allows commercial usage and open source, the 3rd party license is fine.
\nIt is your code (for submission and training), which should be Apache 2.0.

\n
\n

(If the answer to 1. is Yes) Can DKM be used in this competition?
\n I checked the license for DKM, which was frequently used in previous competitions, and it states that only models trained with Megadepth will be released under the MIT license.
\n In the experiments folder of the official repository, it seems that DKMv3_outdoor, which is also used in this Notebook, was trained with Megadepth. So, I am assuming it can be used, is that correct?

\n
\n

That is correct.

\n
\n

am I correct in assuming that SuperGlue cannot be used?
\n Unfortunately, SuperGlue, which was popular in previous competitions, is released under a license that does not allow commercial use. Therefore, I understand that SuperGlue cannot be used in this competition, at least for the official repositories. Is my assumption correct?

\n
\n

That is correct. You also cannot use official SuperPoint, licensed same as SuperGlue.
\nHowever, you can use open sourced SuperPoint https://github.com/rpautrat/SuperPoint, for example.
\nYou might be also interested in this topic https://www.kaggle.com/competitions/image-matching-challenge-2024/discussion/490723

\n

--
\nBest, Dmytro.

", "votes": null, "replies": [ { "id": 2732858, "author_name": "kashiwaba", "author_url": "", "post_date": "04/03/2024 11:18:02", "content": "

I missed that thread, my apologies.
\nThanks for the answer, very clarifying.

", "votes": null, "replies": [ { "id": 2732863, "author_name": "oldufo", "author_url": "", "post_date": "04/03/2024 11:26:08", "content": "

You are welcome. It is nice to see previous challenges winners to participate again :)

", "votes": null, "replies": [] } ] } ] }, { "id": 2790336, "author_name": "sugupoko", "author_url": "", "post_date": "05/03/2024 06:24:29", "content": "

Hi, @oldufo -san!

\n

Is it possible to directly use a model trained on ImageNet that is not for commercial use?
\nFor example, using the timm library to use a model trained on ImageNet and incorporating the results of a 1000-class scene classification into an algorithm, is this allowed?

\n

Using ImageNet for pre-training is often a topic of discussion, but I would like your opinion on the case of direct use.

\n

BR

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2732056": "After reviewing the rules, I noticed that IMC2024 has some differences compared to IMC2022 and IMC2023. It appears that the Winner License this year must be Apache 2.0. Therefore, I have a few questions regarding the license.\n\n1. is the MIT license allowed?\nMy understanding is that Apache 2.0 and MIT are compatible licenses, meaning that software published under the MIT license can also be used. Is my understanding correct?\n\n2. (If the answer to 1. is Yes) Can DKM be used in this competition?\nI checked [the license for DKM](https://github.com/Parskatt/DKM/blob/main/LICENSE), which was frequently used in previous competitions, and it states that only models trained with Megadepth will be released under the MIT license. \nIn the experiments folder of [the official repository](https://github.com/Parskatt/DKM), it seems that DKMv3_outdoor, which is also used in [this Notebook](https://www.kaggle.com/code/motono0223/imagematching2024-dkm-example), was trained with Megadepth. So, I am assuming it can be used, is that correct?\n\n3. am I correct in assuming that SuperGlue cannot be used?\nUnfortunately, [SuperGlue](https://github.com/magicleap/SuperGluePretrainedNetwork), which was popular in previous competitions, is released under a license that does not allow commercial use. Therefore, I understand that SuperGlue cannot be used in this competition, at least for the official repositories. Is my assumption correct?\n\nI would be very grateful if you could answer these questions.", "2732648": "Hi,\n\nI believe, that your questions has been already answered in the Welcome thread: https://www.kaggle.com/competitions/image-matching-challenge-2024/discussion/486584\n\n\n>Prize-eligibility\nAny submission, using non-commercial licensed 3rd party code is not prize eligible. E.g. submissions using SuperGlue (or SuperPoint) are not prize-eligible. Use LightGlue instead, which is both open source, and provides better performance.\nGPL licensed 3rd party code is OK. However, your own code should be Apache 2.0 licensed.\n\n\nHowever, one cannot have too much clarity, so I am happy to answer those again. \nOne of the goals of this competition, is to promote and improve methods, which could be used by anyone freely. If you use this as a guideline principle, the answers are easy to predict:\n\n >is the MIT license allowed?\nMy understanding is that Apache 2.0 and MIT are compatible licenses, meaning that software published under the MIT license can also be used. Is my understanding correct?\n\nYes, you can use any MIT, or other OSS licensed software freely. Even GPLv3, which is not fully compatible with Apache 2.0 is OK. As long, as license allows commercial usage and open source, the 3rd party license is fine. \nIt is your code (for submission and training), which should be Apache 2.0. \n\n>(If the answer to 1. is Yes) Can DKM be used in this competition?\nI checked the license for DKM, which was frequently used in previous competitions, and it states that only models trained with Megadepth will be released under the MIT license.\nIn the experiments folder of the official repository, it seems that DKMv3_outdoor, which is also used in this Notebook, was trained with Megadepth. So, I am assuming it can be used, is that correct?\n\nThat is correct. \n\n>am I correct in assuming that SuperGlue cannot be used?\nUnfortunately, SuperGlue, which was popular in previous competitions, is released under a license that does not allow commercial use. Therefore, I understand that SuperGlue cannot be used in this competition, at least for the official repositories. Is my assumption correct?\n\nThat is correct. You also cannot use official SuperPoint, licensed same as SuperGlue. \nHowever, you can use open sourced SuperPoint https://github.com/rpautrat/SuperPoint, for example.\nYou might be also interested in this topic https://www.kaggle.com/competitions/image-matching-challenge-2024/discussion/490723\n\n\n--\nBest, Dmytro.", "2732858": "I missed that thread, my apologies.\nThanks for the answer, very clarifying.", "2732863": "You are welcome. It is nice to see previous challenges winners to participate again :)", "2790336": "Hi, @oldufo -san!\n\nIs it possible to directly use a model trained on ImageNet that is not for commercial use? \nFor example, using the timm library to use a model trained on ImageNet and incorporating the results of a 1000-class scene classification into an algorithm, is this allowed?\n\nUsing ImageNet for pre-training is often a topic of discussion, but I would like your opinion on the case of direct use.\n\nBR" }, "source": "meta" }