First, we are delighted by the large and active participation in this competition by the Kaggle community.
\n\nThe competition rules state that the winning submissions (data and code) are required to be compliant with the MIT License. This is to ensure that winning submissions are open source, so that they can be used and extended by the worldwide plant phenotyping research community. We chose the MIT License because it is highly non-restrictive. We appreciate that there are alternative open source licenses, as has been discussed in the forums, but at this late stage in the competition we do not feel it would be fair to change the competition rules. You are free to try out any technique you like while crafting your submissions; however the two submissions that you select as your final submissions must be compliant with the competition rules.
\n\nI would also like to remind everyone that this competition is associated with the Computer Vision Problems in Plant Phenotyping workshop at ECCV, and we are interested in advancing novel approaches with academic contributions to solve the wheat head detection task. Approaches used in papers that are submitted to the workshop do not necessarily need to conform to the competition rules, so you are free to use any approach you would like for your academic papers.
\n\nWe would like to thank everyone for the active discussion of this issue on the forums. And good luck in the final weeks of the competition.
\n\nKind regards, \nIan Stavness
", "messages": [ { "id": 923500, "postDate": "2020-07-10T23:29:20.147Z", "content": "First, we are delighted by the large and active participation in this competition by the Kaggle community.
\n\nThe competition rules state that the winning submissions (data and code) are required to be compliant with the MIT License. This is to ensure that winning submissions are open source, so that they can be used and extended by the worldwide plant phenotyping research community. We chose the MIT License because it is highly non-restrictive. We appreciate that there are alternative open source licenses, as has been discussed in the forums, but at this late stage in the competition we do not feel it would be fair to change the competition rules. You are free to try out any technique you like while crafting your submissions; however the two submissions that you select as your final submissions must be compliant with the competition rules.
\n\nI would also like to remind everyone that this competition is associated with the Computer Vision Problems in Plant Phenotyping workshop at ECCV, and we are interested in advancing novel approaches with academic contributions to solve the wheat head detection task. Approaches used in papers that are submitted to the workshop do not necessarily need to conform to the competition rules, so you are free to use any approach you would like for your academic papers.
\n\nWe would like to thank everyone for the active discussion of this issue on the forums. And good luck in the final weeks of the competition.
\n\nKind regards, \nIan Stavness
", "rawMarkdown": "First, we are delighted by the large and active participation in this competition by the Kaggle community.\n\nThe competition rules state that the winning submissions (data and code) are required to be compliant with the MIT License. This is to ensure that winning submissions are open source, so that they can be used and extended by the worldwide plant phenotyping research community. We chose the MIT License because it is highly non-restrictive. We appreciate that there are alternative open source licenses, as has been discussed in the forums, but at this late stage in the competition we do not feel it would be fair to change the competition rules. You are free to try out any technique you like while crafting your submissions; however the two submissions that you select as your final submissions must be compliant with the competition rules.\n\nI would also like to remind everyone that this competition is associated with the Computer Vision Problems in Plant Phenotyping workshop at ECCV, and we are interested in advancing novel approaches with academic contributions to solve the wheat head detection task. Approaches used in papers that are submitted to the workshop do not necessarily need to conform to the competition rules, so you are free to use any approach you would like for your academic papers.\n\nWe would like to thank everyone for the active discussion of this issue on the forums. And good luck in the final weeks of the competition.\n\n\nKind regards, \nIan Stavness", "votes": 20 }, { "id": 923546, "postDate": "2020-07-11T00:55:38.693Z", "content": "Thanks for the update! There's one question that I've seen asked a few places but never answered:\nHow will these rules be enforced across the leaderboard? For example, let's say the first three places all used GPL2. Do they still get to keep their place on the leaderboard? Or will their position be realigned/dropped completely?
\n\nAlso, how far down the leaderboard will you check? I understand that it is not feasible to check every solution for its license but it's also not fair to those who comply with the rules and end up at a lower position because of it.
", "rawMarkdown": "Thanks for the update! There's one question that I've seen asked a few places but never answered:\nHow will these rules be enforced across the leaderboard? For example, let's say the first three places all used GPL2. Do they still get to keep their place on the leaderboard? Or will their position be realigned/dropped completely?\n\nAlso, how far down the leaderboard will you check? I understand that it is not feasible to check every solution for its license but it's also not fair to those who comply with the rules and end up at a lower position because of it.", "votes": 11, "replies": [ { "id": 923571, "postDate": "2020-07-11T01:34:58.867Z", "content": "yeah ，i care more about “How will these rules be enforced across the leaderboard”
", "rawMarkdown": "yeah ，i care more about “How will these rules be enforced across the leaderboard”", "votes": 3 }, { "id": 924719, "postDate": "2020-07-11T15:39:27.723Z", "content": "What I saw on DFDC 1st place removed solution ......\nThe best final submission of this team got the highest LB but was disqualified later because of the rules. And this solution was discarded. But the second solution they submitted followed the competition rules and it gave them 7th place on private LB.\nWill organizers take same actions in this case? \nWill only the submissions not maintaining a proper license be discarded or the whole team will be eliminated?
", "rawMarkdown": "What I saw on [DFDC 1st place removed solution](https://www.kaggle.com/c/deepfake-detection-challenge/discussion/157983) ......\nThe best final submission of this team got the highest LB but was disqualified later because of the rules. And this solution was discarded. But the second solution they submitted followed the competition rules and it gave them 7th place on private LB.\nWill organizers take same actions in this case? \nWill only the submissions not maintaining a proper license be discarded or the whole team will be eliminated?" }, { "id": 928432, "postDate": "2020-07-14T01:44:18.550Z", "content": "The requirement is that the two final submissions must comply with the rules. Individual submissions can be submitted throughout but the two selected must conform.
", "rawMarkdown": "The requirement is that the two final submissions must comply with the rules. Individual submissions can be submitted throughout but the two selected must conform. ", "votes": -4 }, { "id": 928725, "postDate": "2020-07-14T07:16:09.187Z", "content": "Will there be some kind of verificationof what was used for final submittions, because from what I`ve seen 4 days ago new yolov5 notebook was shared and sudennly all the people has 0.77x result. This is a bit unfair to those that are not using yolo
", "rawMarkdown": "Will there be some kind of verificationof what was used for final submittions, because from what I`ve seen 4 days ago new yolov5 notebook was shared and sudennly all the people has 0.77x result. This is a bit unfair to those that are not using yolo ", "votes": 2 }, { "id": 928736, "postDate": "2020-07-14T07:19:44.607Z", "content": "That response was a bit vague and didn't really answer my specific question. I just hope they are able to enforce these rules across the board as there are currently hundreds of people using YoloV5.
", "rawMarkdown": "That response was a bit vague and didn't really answer my specific question. I just hope they are able to enforce these rules across the board as there are currently hundreds of people using YoloV5.", "votes": 8 }, { "id": 955855, "postDate": "2020-08-03T02:28:30.810Z", "content": "Due to the superior performance of YOLOv5, many competitors may still submit the YOLO V5 model as two final submissions or one of the two final submissions。
\n\nSo, may I ask： Will Kaggle check all of the final submissions of all competitors about the license? or Kaggle can only check only a limited number of submissions, such as the top 100, 200?
\n\nAs I know some competitors I recognized want to submit a \"less advanced submissions\" to avoid license check. It's so unfair to only check a small number of submissions.
\n\nHope to get your reply！ Thanks
", "rawMarkdown": "Due to the superior performance of YOLOv5, many competitors may still submit the YOLO V5 model as two final submissions or one of the two final submissions。\n\nSo, may I ask： Will Kaggle check all of the final submissions of all competitors about the license? or Kaggle can only check only a limited number of submissions, such as the top 100, 200?\n\nAs I know some competitors I recognized want to submit a \"less advanced submissions\" to avoid license check. It's so unfair to only check a small number of submissions.\n\nHope to get your reply！ Thanks\n", "votes": 2 } ] }, { "id": 925629, "postDate": "2020-07-12T07:38:35.507Z", "content": "What about we add both the solutions using MIT license and do not end up in the medal zone meanwhile someone with submission with non-MIT license ends up getting the medal.Its seems confusing
", "rawMarkdown": "What about we add both the solutions using MIT license and do not end up in the medal zone meanwhile someone with submission with non-MIT license ends up getting the medal.Its seems confusing", "votes": 10 }, { "id": 924413, "postDate": "2020-07-11T12:15:04.727Z", "content": "Could we use the Apache-2.0 license? (EffDet)
", "rawMarkdown": "Could we use the Apache-2.0 license? (EffDet)", "votes": 7, "replies": [ { "id": 924577, "postDate": "2020-07-11T14:08:36.677Z", "content": "Please look here. The creator of that EffDet implementation gave an in-depth answer on how you can use it.
", "rawMarkdown": "Please look [here](https://www.kaggle.com/c/global-wheat-detection/discussion/164845#923524). The creator of that EffDet implementation gave an in-depth answer on how you can use it.", "votes": 3 }, { "id": 924722, "postDate": "2020-07-11T15:40:39.690Z", "content": "thank you!
", "rawMarkdown": "thank you!" } ] }, { "id": 923612, "postDate": "2020-07-11T02:55:12.670Z", "content": "We are playing game under a set of rules.\nAnd if the rule prohibit the use of GPL2, we should not select the solution which is using GPL2 as our \"final 2 submissions\" regardless of prize as Ian said.\nHost could not be able to check all the final submissions, but this dose not mean we can violate rules.\nIf we were to break the rule intentionally and get a medal, it is called cheating.
\n\nSo I think best way that this competition work is:\n・we select rule-compliant final 2 submissions regardless of prize and medal zone(we dont want to be cheater).\n・Host would check final submissions as many as possible and remove cheaters from LB.
\n\nBut, in my personal opinion, yolov5 should be allowed...\nI hope that hosts will rethinking this rule for greatest happiness of the greatest number.
", "rawMarkdown": "We are playing game under a set of rules.\nAnd if the rule prohibit the use of GPL2, we should not select the solution which is using GPL2 as our \"final 2 submissions\" regardless of prize as Ian said.\nHost could not be able to check all the final submissions, but this dose not mean we can violate rules.\nIf we were to break the rule intentionally and get a medal, it is called cheating.\n\nSo I think best way that this competition work is:\n・we select rule-compliant final 2 submissions regardless of prize and medal zone(we dont want to be cheater).\n・Host would check final submissions as many as possible and remove cheaters from LB.\n\nBut, in my personal opinion, yolov5 should be allowed...\nI hope that hosts will rethinking this rule for greatest happiness of the greatest number.", "votes": 6 }, { "id": 924159, "postDate": "2020-07-11T09:35:13.603Z", "content": "Should I worry the use of the torchvision's license which is BSD 3-Clause?
\n\n\nhowever the two submissions that you select as your final submissions must be compliant with the competition rules.
\n
@stavness wrote this, so I guess everything other than MIT is out of the question now?
", "rawMarkdown": "> however the two submissions that you select as your final submissions must be compliant with the competition rules.\n\n@stavness wrote this, so I guess everything other than MIT is out of the question now?", "votes": 1, "replies": [ { "id": 924475, "postDate": "2020-07-11T12:58:00.683Z", "content": "if this is true, most of the solution in the notebook section are not compliant...\neven torchvision has BSD-3.
", "rawMarkdown": "if this is true, most of the solution in the notebook section are not compliant...\neven torchvision has BSD-3.", "votes": 4 }, { "id": 924609, "postDate": "2020-07-11T14:34:04.603Z", "content": "I read Notebooks and Discussion but I couldn't find out MIT. ( I may have overlooked. )
\n\n| Name | License (URL) |\n| --- | --- |\n| yolov5 | GNU General Public License v3.0 |\n| darknet | YOLO LICENSE|\n| torchvision | BSD 3-Clause \"New\" or \"Revised\" License |\n| efficientdet-pytorch | Apache License 2.0 |\n| mmdetection | Apache License 2.0 |\n| detr | Apache License 2.0|\n| pytorch-lightning | Apache License 2.0|\n| detectron2-ResNeSt | Apache License 2.0|\n| keras-retinanet | Apache License 2.0|\n| pytorch-retinanet | Apache License 2.0|\n| TensorFlow Object Detection API | Apache License 2.0|
", "rawMarkdown": "I read Notebooks and Discussion but I couldn't find out MIT. ( I may have overlooked. )\n\n| Name | License (URL) |\n| --- | --- |\n| yolov5 | [GNU General Public License v3.0 ](https://github.com/ultralytics/yolov5/blob/master/LICENSE)|\n| darknet | [YOLO LICENSE](https://github.com/AlexeyAB/darknet/blob/master/LICENSE)|\n| torchvision | [BSD 3-Clause \"New\" or \"Revised\" License ](https://github.com/pytorch/vision/blob/master/LICENSE)|\n| efficientdet-pytorch | [Apache License 2.0 ](https://github.com/rwightman/efficientdet-pytorch/blob/master/LICENSE)|\n| mmdetection | [Apache License 2.0 ](https://github.com/open-mmlab/mmdetection/blob/master/LICENSE)|\n| detr | [Apache License 2.0](https://github.com/facebookresearch/detr/blob/master/LICENSE)|\n| pytorch-lightning | [Apache License 2.0](https://github.com/PyTorchLightning/pytorch-lightning/blob/master/LICENSE)|\n| detectron2-ResNeSt | [Apache License 2.0](https://github.com/zhanghang1989/detectron2-ResNeSt/blob/resnest/LICENSE)|\n| keras-retinanet | [Apache License 2.0](https://github.com/fizyr/keras-retinanet/blob/master/LICENSE)|\n| pytorch-retinanet | [Apache License 2.0](https://github.com/yhenon/pytorch-retinanet/blob/master/LICENSE)|\n| TensorFlow Object Detection API | [Apache License 2.0](https://github.com/tensorflow/models/blob/master/LICENSE)|\n\n\n", "votes": 20 }, { "id": 925450, "postDate": "2020-07-12T05:20:10.397Z", "content": "@nakajima Hi there, models that under Apache Licence are considered \"compatible with MIT License\", is that correct? Thanks!
", "rawMarkdown": "@nakajima Hi there, models that under Apache Licence are considered \"compatible with MIT License\", is that correct? Thanks!" }, { "id": 925463, "postDate": "2020-07-12T05:26:09.743Z", "content": "Look here. Ross, who wrote the efficientdet pytorch implementation gives an in-depth explanation on how you can use it.
", "rawMarkdown": "Look [here](https://www.kaggle.com/c/global-wheat-detection/discussion/164845#923524). Ross, who wrote the efficientdet pytorch implementation gives an in-depth explanation on how you can use it.", "votes": 1 }, { "id": 940476, "postDate": "2020-07-23T01:15:36.367Z", "content": "I think it would be very beneficial and fair if the host provide a list of acceptable / unacceptable licenses for popular models as tn did above.
", "rawMarkdown": "I think it would be very beneficial and fair if the host provide a list of acceptable / unacceptable licenses for popular models as tn did above.", "votes": -1 } ] }, { "id": 925368, "postDate": "2020-07-12T03:24:41.697Z", "content": "If the final two solutions one is using non-MIT license, but another one is using MIT license, the non-MIT one can get the medal and the MIT can try to be in the money?
No, like it says in the post
\n\n\n", "rawMarkdown": "No, like it says in the post\n> the two submissions that you select as your final submissions must be compliant with the competition rules", "votes": 3 } ] }, { "id": 923664, "postDate": "2020-07-11T04:23:41.213Z", "content": "the two submissions that you select as your final submissions must be compliant with the competition rules
\n
Thank you for your support, this means a lot for me!
", "rawMarkdown": "Thank you for your support, this means a lot for me!" }, { "id": 930363, "postDate": "2020-07-15T11:55:46.797Z", "content": "This would also mean YOLOv3 which is by the same author as YOLOv5\nhttps://github.com/ultralytics/yolov3/blob/master/LICENSE\nGNU General Public License v3.0
", "rawMarkdown": "This would also mean YOLOv3 which is by the same author as YOLOv5\nhttps://github.com/ultralytics/yolov3/blob/master/LICENSE\nGNU General Public License v3.0", "votes": -5 }, { "id": 926836, "postDate": "2020-07-13T03:14:44.370Z", "content": "What about only importing the library(GLP-v3) and not modifying it, can this fit the rule?
", "rawMarkdown": "What about only importing the library(GLP-v3) and not modifying it, can this fit the rule?", "votes": -1, "replies": [ { "id": 929632, "postDate": "2020-07-14T20:04:09.137Z", "content": "Under the typical interpretations of GPL that I'm familiar with, you'd be allowed to import an LGPL licensed module (without modifying or including the source) and keep an MIT license on your solution but you would not be allowed to import a GPLv3 module without also licensing your solution as GPLv3.
", "rawMarkdown": "Under the typical interpretations of GPL that I'm familiar with, you'd be allowed to import an LGPL licensed module (without modifying or including the source) and keep an MIT license on your solution but you would not be allowed to import a GPLv3 module without also licensing your solution as GPLv3.", "votes": 2 } ] }, { "id": 958481, "postDate": "2020-08-05T02:42:36.380Z", "content": "YOLOv5 does not conform to the rules of the competition and should be removed from the leaderboard?
", "rawMarkdown": " YOLOv5 does not conform to the rules of the competition and should be removed from the leaderboard?" }, { "id": 923501, "postDate": "2020-07-10T23:33:12.160Z", "content": "Would the solutions that are using non-MIT license be invalidated regardless of the position on the leaderboard?
", "rawMarkdown": "Would the solutions that are using non-MIT license be invalidated regardless of the position on the leaderboard?", "replies": [ { "id": 923504, "postDate": "2020-07-10T23:36:14.393Z", "content": "Solutions that are using a non-MIT license do not comply with the rules to win the competition.
", "rawMarkdown": "Solutions that are using a non-MIT license do not comply with the rules to win the competition. ", "votes": 1 }, { "id": 923508, "postDate": "2020-07-10T23:43:22.647Z", "rawMarkdown": "", "isDeleted": true }, { "id": 923577, "postDate": "2020-07-11T01:50:49.573Z", "content": "What is the win for the competition? Does it mean in the money or all the medals are also win?
@maggiemd And just to be clear, is it possible to use code whose lisence is not MIT, but MIT-compatible?
", "rawMarkdown": "@maggiemd And just to be clear, is it possible to use code whose lisence is not MIT, but MIT-compatible?", "votes": 4 }, { "id": 928434, "postDate": "2020-07-14T01:46:13.483Z", "content": "As with any competition, if we become aware that someone had not conformed with the rules then the host has the discretion to disqualify.
", "rawMarkdown": "As with any competition, if we become aware that someone had not conformed with the rules then the host has the discretion to disqualify. ", "votes": -3 }, { "id": 929407, "postDate": "2020-07-14T16:46:24.247Z", "content": "Dear @maggiemd, @stavness, @bendvd, I think it is super obvious now that people are not understanding the rules anymore. Can you *please * post another competition update/clarification, clearly stating what's going on? \nWill all yolov5 based final submissions be removed? Or is it only impossible to WIN (=in the money) with yolov5?\nI invested a huge amout of time and energy (literally, ask my GPU!) in yolov5 and so did others. After all the other ...suboptimal developments in this competition, please clarifiy this to avoid frustration for hundreds of competition colleagues here.
", "rawMarkdown": "Dear @maggiemd, @stavness, @bendvd, I think it is super obvious now that people are not understanding the rules anymore. Can you *please * post another competition update/clarification, clearly stating what's going on? \nWill all yolov5 based final submissions be removed? Or is it only impossible to WIN (=in the money) with yolov5?\nI invested a huge amout of time and energy (literally, ask my GPU!) in yolov5 and so did others. After all the other ...suboptimal developments in this competition, please clarifiy this to avoid frustration for hundreds of competition colleagues here.", "votes": 5 }, { "id": 935029, "postDate": "2020-07-19T03:16:01.527Z", "content": "Hi @maggiemd , I'm a little confused. I have read the rules of this competition, in the section A. COMPETITION-SPECIFIC RULES, 1. WINNER LICENSE., it said the licenses that in no event limits commercial use are OK.
\nOpen Source: You hereby license and will license your winning Submission and the source code used to generate the Submission under an Open Source Initiative-approved license (see www.opensource.org) that in no event limits commercial use of such code or model containing or depending on such code. For generally commercially available software that you used to generate your submission that is not owned by you, but that can be procured by the Competition Sponsor without undue expense, you do not need to grant the license in the preceding sentence for that software.\n
So I licensed my training script with Apache-2.0 License, it's a license that in no event limits commercial use. So I think it complies with the rules of this competition.
\n\nBut here, you said Solutions that are using a non-MIT license do not comply with the rules. Could you explain it to me, please?
@ufownl Take a look at the newer update from 3 days ago which states the following:
\n\n\n\n\nThere are many commonly used permissive open source licenses (e.g., Apache 2.0, BSD 3.0, others) that are compatible with the MIT License and therefore are compliant with the Competition Rules
\n
So yes, Apache-2.0 does fit the licensing requirement. Take a look at Ross's comment here on how you should include the licence in your submission.
", "rawMarkdown": "@ufownl Take a look at the [newer update from 3 days ago](https://www.kaggle.com/c/global-wheat-detection/discussion/167298) which states the following:\n> There are many commonly used permissive open source licenses (e.g., Apache 2.0, BSD 3.0, others) that are compatible with the MIT License and therefore are compliant with the Competition Rules\n\nSo yes, Apache-2.0 does fit the licensing requirement. Take a look at Ross's comment [here](https://www.kaggle.com/c/global-wheat-detection/discussion/164845#923524) on how you should include the licence in your submission.", "votes": 1 }, { "id": 935058, "postDate": "2020-07-19T03:59:48.953Z", "content": "@matthewmasters Thanks for your explanation. 👍
", "rawMarkdown": "@matthewmasters Thanks for your explanation. 👍 " } ] }, { "id": 927680, "postDate": "2020-07-13T14:39:04.797Z", "rawMarkdown": "", "isDeleted": true }, { "id": 928015, "postDate": "2020-07-13T17:28:51.120Z", "content": "Thank you all for clear explanation
", "rawMarkdown": "Thank you all for clear explanation", "votes": 1 } ], "comments": [ { "id": 923546, "author_name": "Matt", "author_url": "", "post_date": "2020-07-11T00:55:38.693000", "content": "Thanks for the update! There's one question that I've seen asked a few places but never answered:\nHow will these rules be enforced across the leaderboard? For example, let's say the first three places all used GPL2. Do they still get to keep their place on the leaderboard? Or will their position be realigned/dropped completely?
\n\nAlso, how far down the leaderboard will you check? I understand that it is not feasible to check every solution for its license but it's also not fair to those who comply with the rules and end up at a lower position because of it.
", "votes": 11, "replies": [ { "id": 923571, "author_name": "天宇_tianyu", "author_url": "", "post_date": "2020-07-11T01:34:58.867000", "content": "yeah ，i care more about “How will these rules be enforced across the leaderboard”
", "votes": 3, "replies": [] }, { "id": 924719, "author_name": "Zaber Ibn Abdul Hakim", "author_url": "", "post_date": "2020-07-11T15:39:27.723000", "content": "What I saw on DFDC 1st place removed solution ......\nThe best final submission of this team got the highest LB but was disqualified later because of the rules. And this solution was discarded. But the second solution they submitted followed the competition rules and it gave them 7th place on private LB.\nWill organizers take same actions in this case? \nWill only the submissions not maintaining a proper license be discarded or the whole team will be eliminated?
", "votes": 0, "replies": [] }, { "id": 928432, "author_name": "Maggie", "author_url": "", "post_date": "2020-07-14T01:44:18.550000", "content": "The requirement is that the two final submissions must comply with the rules. Individual submissions can be submitted throughout but the two selected must conform.
", "votes": -4, "replies": [] }, { "id": 928725, "author_name": "Marcin Czelej", "author_url": "", "post_date": "2020-07-14T07:16:09.187000", "content": "Will there be some kind of verificationof what was used for final submittions, because from what I`ve seen 4 days ago new yolov5 notebook was shared and sudennly all the people has 0.77x result. This is a bit unfair to those that are not using yolo
", "votes": 2, "replies": [] }, { "id": 928736, "author_name": "Matt", "author_url": "", "post_date": "2020-07-14T07:19:44.607000", "content": "That response was a bit vague and didn't really answer my specific question. I just hope they are able to enforce these rules across the board as there are currently hundreds of people using YoloV5.
", "votes": 8, "replies": [] }, { "id": 955855, "author_name": "StoneWST98", "author_url": "", "post_date": "2020-08-03T02:28:30.810000", "content": "Due to the superior performance of YOLOv5, many competitors may still submit the YOLO V5 model as two final submissions or one of the two final submissions。
\n\nSo, may I ask： Will Kaggle check all of the final submissions of all competitors about the license? or Kaggle can only check only a limited number of submissions, such as the top 100, 200?
\n\nAs I know some competitors I recognized want to submit a \"less advanced submissions\" to avoid license check. It's so unfair to only check a small number of submissions.
\n\nHope to get your reply！ Thanks
", "votes": 2, "replies": [] } ] }, { "id": 925629, "author_name": "AbhinavDutt", "author_url": "", "post_date": "2020-07-12T07:38:35.507000", "content": "What about we add both the solutions using MIT license and do not end up in the medal zone meanwhile someone with submission with non-MIT license ends up getting the medal.Its seems confusing
", "votes": 10, "replies": [] }, { "id": 924413, "author_name": "Wonjun Park", "author_url": "", "post_date": "2020-07-11T12:15:04.727000", "content": "Could we use the Apache-2.0 license? (EffDet)
", "votes": 7, "replies": [ { "id": 924577, "author_name": "Matt", "author_url": "", "post_date": "2020-07-11T14:08:36.677000", "content": "Please look here. The creator of that EffDet implementation gave an in-depth answer on how you can use it.
", "votes": 3, "replies": [] }, { "id": 924722, "author_name": "Wonjun Park", "author_url": "", "post_date": "2020-07-11T15:40:39.690000", "content": "thank you!
", "votes": 0, "replies": [] } ] }, { "id": 923612, "author_name": "tito", "author_url": "", "post_date": "2020-07-11T02:55:12.670000", "content": "We are playing game under a set of rules.\nAnd if the rule prohibit the use of GPL2, we should not select the solution which is using GPL2 as our \"final 2 submissions\" regardless of prize as Ian said.\nHost could not be able to check all the final submissions, but this dose not mean we can violate rules.\nIf we were to break the rule intentionally and get a medal, it is called cheating.
\n\nSo I think best way that this competition work is:\n・we select rule-compliant final 2 submissions regardless of prize and medal zone(we dont want to be cheater).\n・Host would check final submissions as many as possible and remove cheaters from LB.
\n\nBut, in my personal opinion, yolov5 should be allowed...\nI hope that hosts will rethinking this rule for greatest happiness of the greatest number.
", "votes": 6, "replies": [] }, { "id": 924159, "author_name": "Chanran Kim", "author_url": "", "post_date": "2020-07-11T09:35:13.603000", "content": "Should I worry the use of the torchvision's license which is BSD 3-Clause?
\n\n\nhowever the two submissions that you select as your final submissions must be compliant with the competition rules.
\n
@stavness wrote this, so I guess everything other than MIT is out of the question now?
", "votes": 1, "replies": [ { "id": 924475, "author_name": "Geno87", "author_url": "", "post_date": "2020-07-11T12:58:00.683000", "content": "if this is true, most of the solution in the notebook section are not compliant...\neven torchvision has BSD-3.
", "votes": 4, "replies": [] }, { "id": 924609, "author_name": "tn", "author_url": "", "post_date": "2020-07-11T14:34:04.603000", "content": "I read Notebooks and Discussion but I couldn't find out MIT. ( I may have overlooked. )
\n\n| Name | License (URL) |\n| --- | --- |\n| yolov5 | GNU General Public License v3.0 |\n| darknet | YOLO LICENSE|\n| torchvision | BSD 3-Clause \"New\" or \"Revised\" License |\n| efficientdet-pytorch | Apache License 2.0 |\n| mmdetection | Apache License 2.0 |\n| detr | Apache License 2.0|\n| pytorch-lightning | Apache License 2.0|\n| detectron2-ResNeSt | Apache License 2.0|\n| keras-retinanet | Apache License 2.0|\n| pytorch-retinanet | Apache License 2.0|\n| TensorFlow Object Detection API | Apache License 2.0|
", "votes": 20, "replies": [] }, { "id": 925450, "author_name": "", "author_url": "", "post_date": "2020-07-12T05:20:10.397000", "content": "@nakajima Hi there, models that under Apache Licence are considered \"compatible with MIT License\", is that correct? Thanks!
", "votes": 0, "replies": [] }, { "id": 925463, "author_name": "Matt", "author_url": "", "post_date": "2020-07-12T05:26:09.743000", "content": "Look here. Ross, who wrote the efficientdet pytorch implementation gives an in-depth explanation on how you can use it.
", "votes": 1, "replies": [] }, { "id": 940476, "author_name": "kuroyuli", "author_url": "", "post_date": "2020-07-23T01:15:36.367000", "content": "I think it would be very beneficial and fair if the host provide a list of acceptable / unacceptable licenses for popular models as tn did above.
", "votes": -1, "replies": [] } ] }, { "id": 925368, "author_name": "Zhongxuan Luo", "author_url": "", "post_date": "2020-07-12T03:24:41.697000", "content": "If the final two solutions one is using non-MIT license, but another one is using MIT license, the non-MIT one can get the medal and the MIT can try to be in the money?
No, like it says in the post
\n\n\n", "votes": 3, "replies": [] } ] }, { "id": 923664, "author_name": "Peiyuan Liao", "author_url": "", "post_date": "2020-07-11T04:23:41.213000", "content": "the two submissions that you select as your final submissions must be compliant with the competition rules
\n
Thank you for your support, this means a lot for me!
", "votes": 0, "replies": [] }, { "id": 930363, "author_name": "Charlie Craine", "author_url": "", "post_date": "2020-07-15T11:55:46.797000", "content": "This would also mean YOLOv3 which is by the same author as YOLOv5\nhttps://github.com/ultralytics/yolov3/blob/master/LICENSE\nGNU General Public License v3.0
", "votes": -5, "replies": [] }, { "id": 926836, "author_name": "cxt", "author_url": "", "post_date": "2020-07-13T03:14:44.370000", "content": "What about only importing the library(GLP-v3) and not modifying it, can this fit the rule?
", "votes": -1, "replies": [ { "id": 929632, "author_name": "RossWightman", "author_url": "", "post_date": "2020-07-14T20:04:09.137000", "content": "Under the typical interpretations of GPL that I'm familiar with, you'd be allowed to import an LGPL licensed module (without modifying or including the source) and keep an MIT license on your solution but you would not be allowed to import a GPLv3 module without also licensing your solution as GPLv3.
", "votes": 2, "replies": [] } ] }, { "id": 958481, "author_name": "Monster-H7", "author_url": "", "post_date": "2020-08-05T02:42:36.380000", "content": "YOLOv5 does not conform to the rules of the competition and should be removed from the leaderboard?
", "votes": 0, "replies": [] }, { "id": 923501, "author_name": "johnny", "author_url": "", "post_date": "2020-07-10T23:33:12.160000", "content": "Would the solutions that are using non-MIT license be invalidated regardless of the position on the leaderboard?
", "votes": 0, "replies": [ { "id": 923504, "author_name": "Maggie", "author_url": "", "post_date": "2020-07-10T23:36:14.393000", "content": "Solutions that are using a non-MIT license do not comply with the rules to win the competition.
", "votes": 1, "replies": [] }, { "id": 923508, "author_name": "", "author_url": "", "post_date": "2020-07-10T23:43:22.647000", "content": "", "votes": 0, "replies": [] }, { "id": 923577, "author_name": "XY", "author_url": "", "post_date": "2020-07-11T01:50:49.573000", "content": "What is the win for the competition? Does it mean in the money or all the medals are also win?
@maggiemd And just to be clear, is it possible to use code whose lisence is not MIT, but MIT-compatible?
", "votes": 4, "replies": [] }, { "id": 928434, "author_name": "Maggie", "author_url": "", "post_date": "2020-07-14T01:46:13.483000", "content": "As with any competition, if we become aware that someone had not conformed with the rules then the host has the discretion to disqualify.
", "votes": -3, "replies": [] }, { "id": 929407, "author_name": "from coffee import *", "author_url": "", "post_date": "2020-07-14T16:46:24.247000", "content": "Dear @maggiemd, @stavness, @bendvd, I think it is super obvious now that people are not understanding the rules anymore. Can you *please * post another competition update/clarification, clearly stating what's going on? \nWill all yolov5 based final submissions be removed? Or is it only impossible to WIN (=in the money) with yolov5?\nI invested a huge amout of time and energy (literally, ask my GPU!) in yolov5 and so did others. After all the other ...suboptimal developments in this competition, please clarifiy this to avoid frustration for hundreds of competition colleagues here.
", "votes": 5, "replies": [] }, { "id": 935029, "author_name": "RangerUFO", "author_url": "", "post_date": "2020-07-19T03:16:01.527000", "content": "Hi @maggiemd , I'm a little confused. I have read the rules of this competition, in the section A. COMPETITION-SPECIFIC RULES, 1. WINNER LICENSE., it said the licenses that in no event limits commercial use are OK.
\nOpen Source: You hereby license and will license your winning Submission and the source code used to generate the Submission under an Open Source Initiative-approved license (see www.opensource.org) that in no event limits commercial use of such code or model containing or depending on such code. For generally commercially available software that you used to generate your submission that is not owned by you, but that can be procured by the Competition Sponsor without undue expense, you do not need to grant the license in the preceding sentence for that software.\n
So I licensed my training script with Apache-2.0 License, it's a license that in no event limits commercial use. So I think it complies with the rules of this competition.
\n\nBut here, you said Solutions that are using a non-MIT license do not comply with the rules. Could you explain it to me, please?
@ufownl Take a look at the newer update from 3 days ago which states the following:
\n\n\n\n\nThere are many commonly used permissive open source licenses (e.g., Apache 2.0, BSD 3.0, others) that are compatible with the MIT License and therefore are compliant with the Competition Rules
\n
So yes, Apache-2.0 does fit the licensing requirement. Take a look at Ross's comment here on how you should include the licence in your submission.
", "votes": 1, "replies": [] }, { "id": 935058, "author_name": "RangerUFO", "author_url": "", "post_date": "2020-07-19T03:59:48.953000", "content": "@matthewmasters Thanks for your explanation. 👍
", "votes": 0, "replies": [] } ] }, { "id": 927680, "author_name": "", "author_url": "", "post_date": "2020-07-13T14:39:04.797000", "content": "", "votes": 0, "replies": [] }, { "id": 928015, "author_name": "Erdem Erbaba", "author_url": "", "post_date": "2020-07-13T17:28:51.120000", "content": "Thank you all for clear explanation
", "votes": 1, "replies": [] } ], "raw_markdown_by_id": { "923500": "First, we are delighted by the large and active participation in this competition by the Kaggle community.\n\nThe competition rules state that the winning submissions (data and code) are required to be compliant with the MIT License. This is to ensure that winning submissions are open source, so that they can be used and extended by the worldwide plant phenotyping research community. We chose the MIT License because it is highly non-restrictive. We appreciate that there are alternative open source licenses, as has been discussed in the forums, but at this late stage in the competition we do not feel it would be fair to change the competition rules. You are free to try out any technique you like while crafting your submissions; however the two submissions that you select as your final submissions must be compliant with the competition rules.\n\nI would also like to remind everyone that this competition is associated with the Computer Vision Problems in Plant Phenotyping workshop at ECCV, and we are interested in advancing novel approaches with academic contributions to solve the wheat head detection task. Approaches used in papers that are submitted to the workshop do not necessarily need to conform to the competition rules, so you are free to use any approach you would like for your academic papers.\n\nWe would like to thank everyone for the active discussion of this issue on the forums. And good luck in the final weeks of the competition.\n\n\nKind regards, \nIan Stavness", "923546": "Thanks for the update! There's one question that I've seen asked a few places but never answered:\nHow will these rules be enforced across the leaderboard? For example, let's say the first three places all used GPL2. Do they still get to keep their place on the leaderboard? Or will their position be realigned/dropped completely?\n\nAlso, how far down the leaderboard will you check? I understand that it is not feasible to check every solution for its license but it's also not fair to those who comply with the rules and end up at a lower position because of it.", "925629": "What about we add both the solutions using MIT license and do not end up in the medal zone meanwhile someone with submission with non-MIT license ends up getting the medal.Its seems confusing", "924413": "Could we use the Apache-2.0 license? (EffDet)", "923612": "We are playing game under a set of rules.\nAnd if the rule prohibit the use of GPL2, we should not select the solution which is using GPL2 as our \"final 2 submissions\" regardless of prize as Ian said.\nHost could not be able to check all the final submissions, but this dose not mean we can violate rules.\nIf we were to break the rule intentionally and get a medal, it is called cheating.\n\nSo I think best way that this competition work is:\n・we select rule-compliant final 2 submissions regardless of prize and medal zone(we dont want to be cheater).\n・Host would check final submissions as many as possible and remove cheaters from LB.\n\nBut, in my personal opinion, yolov5 should be allowed...\nI hope that hosts will rethinking this rule for greatest happiness of the greatest number.", "924159": "Should I worry the use of the `torchvision`'s license which is BSD 3-Clause?", "923606": "> however the two submissions that you select as your final submissions must be compliant with the competition rules.\n\n@stavness wrote this, so I guess everything other than MIT is out of the question now?", "925368": "If the final two solutions one is using non-MIT license, but another one is using MIT license, the non-MIT one can `get the medal` and the MIT can try to be `in the money`?", "923664": "Thank you for your support, this means a lot for me!", "930363": "This would also mean YOLOv3 which is by the same author as YOLOv5\nhttps://github.com/ultralytics/yolov3/blob/master/LICENSE\nGNU General Public License v3.0", "926836": "What about only importing the library(GLP-v3) and not modifying it, can this fit the rule?", "958481": " YOLOv5 does not conform to the rules of the competition and should be removed from the leaderboard?", "923501": "Would the solutions that are using non-MIT license be invalidated regardless of the position on the leaderboard?", "927680": "", "928015": "Thank you all for clear explanation" } }