As all YoloV5 solutions are now invalidated because of the incompatible license. But as of now, they are working best (based on the leaderboard, ~almost all top 50 teams use it) so far. It's also mentioned by competition host that only winning solutions (top 3) have to be MIT licensed but the license of solutions in the medal zone doesn't matter much.
\n\nNow, having said that (and assuming Yolov5 will be the best till the end), here is the paradox: To win the competition, you will have to use Yolov5 (which will be invalidated if you actually win). If you are using techniques other than Yolov5 but that doesn't score as high as Yolov5, you won't be in the top three in the first place (even if your solution is second best after Yolov5 and eligible to win) because of Yolov5 submissions by the guys who want to be in medal zone (actually everyone who uses Yolov5, because who doesn't want medals, right?) will be there ahead of you in the leaderboard.
\n\nThere are some assumptions in this (like yolov5 is best and no other MIT licensed approach gives scores as high as Yolov5), but they are valid as per the current situation of the competition.
\n\n@maggiemd , @bendvd What do you think about this?
", "messages": [ { "id": 921264, "postDate": "2020-07-09T07:02:39.800Z", "content": "As all YoloV5 solutions are now invalidated because of the incompatible license. But as of now, they are working best (based on the leaderboard, ~almost all top 50 teams use it) so far. It's also mentioned by competition host that only winning solutions (top 3) have to be MIT licensed but the license of solutions in the medal zone doesn't matter much.
\n\nNow, having said that (and assuming Yolov5 will be the best till the end), here is the paradox: To win the competition, you will have to use Yolov5 (which will be invalidated if you actually win). If you are using techniques other than Yolov5 but that doesn't score as high as Yolov5, you won't be in the top three in the first place (even if your solution is second best after Yolov5 and eligible to win) because of Yolov5 submissions by the guys who want to be in medal zone (actually everyone who uses Yolov5, because who doesn't want medals, right?) will be there ahead of you in the leaderboard.
\n\nThere are some assumptions in this (like yolov5 is best and no other MIT licensed approach gives scores as high as Yolov5), but they are valid as per the current situation of the competition.
\n\n@maggiemd , @bendvd What do you think about this?
", "rawMarkdown": "As all YoloV5 solutions are now invalidated because of the incompatible license. But as of now, they are working best (based on the leaderboard, ~almost all top 50 teams use it) **so far**. It's also mentioned by competition host that only winning solutions (top 3) have to be MIT licensed but the license of solutions in the medal zone doesn't matter much.\n\nNow, having said that (and assuming **Yolov5 will be the best till the end**), here is the paradox: To win the competition, you will have to use Yolov5 (which will be invalidated if you actually win). If you are using techniques other than Yolov5 but that doesn't score as high as Yolov5, you won't be in the top three in the first place (even if your solution is second best after Yolov5 and eligible to win) because of Yolov5 submissions by the guys who want to be in medal zone (actually everyone who uses Yolov5, because who doesn't want medals, right?) will be there ahead of you in the leaderboard.\n\nThere are some assumptions in this (like yolov5 is best and no other MIT licensed approach gives scores as high as Yolov5), but they are valid as per the current situation of the competition.\n\n@maggiemd , @bendvd What do you think about this?", "votes": 12 }, { "id": 921430, "postDate": "2020-07-09T09:14:22.170Z", "content": "Also, I'm still confused about a point I brought up in another thread, which is as follows: say hypothetically one of the top three teams ended up using Yolov5 or other code not permitted by the competition. Do they move on to the next team up on the leaderboard to check for eligibility and straight up delete the non-compliant team? What if this goes on for a while and potentially lots of teams get disqualified because they (originally was only in say the gold medal zone) moved to the \"in-the-money\" zone (because the teams before them got deleted) and get checked for eligibility? I'm looking for a formal answer from the organizer because the answer to this affects which pipeline I choose to develop in the following weeks. Yolov5 indeed works well but there are other strong-performing models that I developed. Thanks!
", "rawMarkdown": "Also, I'm still confused about a point I brought up in another thread, which is as follows: say hypothetically one of the top three teams ended up using Yolov5 or other code not permitted by the competition. Do they move on to the next team up on the leaderboard to check for eligibility and straight up delete the non-compliant team? What if this goes on for a while and potentially lots of teams get disqualified because they (originally was only in say the gold medal zone) moved to the \"in-the-money\" zone (because the teams before them got deleted) and get checked for eligibility? I'm looking for a formal answer from the organizer because the answer to this affects which pipeline I choose to develop in the following weeks. Yolov5 indeed works well but there are other strong-performing models that I developed. Thanks!", "votes": 9 }, { "id": 921413, "postDate": "2020-07-09T08:56:14.227Z", "content": "This is a good point, I've also been considering this. Possibly on this occasion Kaggle could check all gold medal winning entries not just the prize zone? Otherwise it feels pretty unfair to people who do stick to the clarified rules. Hopefully other solutions will prove to be equally effective as Etienne suggests. Also his comment makes it sound like a shakeup is on the way as private and public test sets are different.
", "rawMarkdown": "This is a good point, I've also been considering this. Possibly on this occasion Kaggle could check all gold medal winning entries not just the prize zone? Otherwise it feels pretty unfair to people who do stick to the clarified rules. Hopefully other solutions will prove to be equally effective as Etienne suggests. Also his comment makes it sound like a shakeup is on the way as private and public test sets are different.", "votes": 3 }, { "id": 921270, "postDate": "2020-07-09T07:08:05.887Z", "content": "It is a possible situation, but there is still few days before the end of the competition. The private test set is also different from the public test set and will be corrected, it can also shake up the final leaderboard !
", "rawMarkdown": "It is a possible situation, but there is still few days before the end of the competition. The private test set is also different from the public test set and will be corrected, it can also shake up the final leaderboard !", "votes": 2 }, { "id": 921554, "postDate": "2020-07-09T11:27:38.043Z", "content": "In case the 3 teams at \"in the money\" gets disqualifed, that means the 4th-6th teams will climb to 1st-3rd places?\nBecause if thats true, the 4th-6th teams will also gets disqualifed (probably), and if we continue on that way everyone at the top of LB will get disqualifed untill we find the first team who did not use Yolov5.
\n\nSo yeah there is a paradox here, because if I use yolov5 I want to get higher score as I can, but on the other hand I should hope to get lower score than the real winners (who did not use Yolov5) so I wont get disqualifed.
\n\nI think the best solution is to not disqualify anyone, and give the money prizes for the higest score who stand in the competition rules (which will probably be at 20+ at LB).
", "rawMarkdown": "In case the 3 teams at \"in the money\" gets disqualifed, that means the 4th-6th teams will climb to 1st-3rd places?\nBecause if thats true, the 4th-6th teams will also gets disqualifed (probably), and if we continue on that way everyone at the top of LB will get disqualifed untill we find the first team who did not use Yolov5.\n\nSo yeah there is a paradox here, because if I use yolov5 I want to get higher score as I can, but on the other hand I should hope to get lower score than the real winners (who did not use Yolov5) so I wont get disqualifed.\n\nI think the best solution is to not disqualify anyone, and give the money prizes for the higest score who stand in the competition rules (which will probably be at 20+ at LB).", "votes": 2, "replies": [ { "id": 921618, "postDate": "2020-07-09T12:27:34.917Z", "content": "Well, the problem for that is some of the competitors care more about the rank than the prize money, and if they complied with the rules from the beginning, that would be unfair for them.
", "rawMarkdown": "Well, the problem for that is some of the competitors care more about the rank than the prize money, and if they complied with the rules from the beginning, that would be unfair for them.", "votes": 4 }, { "id": 921695, "postDate": "2020-07-09T13:37:43.753Z", "content": "Yes I completely agree, for some people the rank and especially a gold medal is the main motivation.
", "rawMarkdown": "Yes I completely agree, for some people the rank and especially a gold medal is the main motivation. ", "votes": 2 } ] }, { "id": 922361, "postDate": "2020-07-10T03:52:46.017Z", "content": "The License rule is clarified just a few days ago, and right in the middle of the competition period time. \nTo me, it is quite unfair for all people participate here, because they just did their part, yolov5 or not, everyone must pick up something to begin, no one can compete from building a framework.\nSo, it is clear the host's fault and responsibility, they should anticipate this situation, and they can make this right!\nThe host need only make an enquiry to YOLOV5 owner to get a license, pay some money to get the highest performance.
", "rawMarkdown": "The License rule is clarified just a few days ago, and right in the middle of the competition period time. \nTo me, it is quite unfair for all people participate here, because they just did their part, yolov5 or not, everyone must pick up something to begin, no one can compete from building a framework.\nSo, it is clear the host's fault and responsibility, they should anticipate this situation, and they can make this right!\nThe host need only make an enquiry to YOLOV5 owner to get a license, pay some money to get the highest performance. ", "votes": 1, "replies": [ { "id": 922535, "postDate": "2020-07-10T07:23:46.927Z", "content": "In my opinion, it wouldn't be fair to allow it at this point. That would amount to changing the rules which we all agreed to when joining the competition. It would be unfair to the people who abided by the licencing rule and have already made submissions that fit the rule. I agree it's not the best situation, but the best thing to do is learn our lesson and not repeat our mistakes in the future.
", "rawMarkdown": "In my opinion, it wouldn't be fair to allow it at this point. That would amount to changing the rules which we all agreed to when joining the competition. It would be unfair to the people who abided by the licencing rule and have already made submissions that fit the rule. I agree it's not the best situation, but the best thing to do is learn our lesson and not repeat our mistakes in the future.", "votes": 1 }, { "id": 922626, "postDate": "2020-07-10T08:37:24.787Z", "content": "The ridiculousness of this rule is like to ban BERT for an NLP competition, rule out pretrained weights for any CV competition, not allowing Pi doing a competition on quadrature a circle .rule out black people from doing something
\n\nNot every rule is a good rule ,thinking of Hitler's Nurnberg Laws. sometimes, it need speople to disobey the rules to make things right!!!!
", "rawMarkdown": "The ridiculousness of this rule is like to ban BERT for an NLP competition, rule out pretrained weights for any CV competition, not allowing Pi doing a competition on quadrature a circle .rule out black people from doing something\n\nNot every rule is a good rule ,thinking of Hitler's Nurnberg Laws. sometimes, it need speople to disobey the rules to make things right!!!!\n\n\n", "votes": -6 }, { "id": 922700, "postDate": "2020-07-10T09:25:33.280Z", "content": "It's okay to disagree with my opinion and downvote my comment as you have. But comparing the Kaggle rules to racism and the Nuremberg laws is a little too far. Just remember the Kaggle etiquette of \"Be patient, be friendly, and focus on ideas\".
", "rawMarkdown": "It's okay to disagree with my opinion and downvote my comment as you have. But comparing the Kaggle rules to racism and the Nuremberg laws is a little too far. Just remember the Kaggle etiquette of \"Be patient, be friendly, and focus on ideas\".", "votes": 6 }, { "id": 922916, "postDate": "2020-07-10T12:16:55.510Z", "content": "I didn't downvote your reply. I welcome you to comment on my thread,and I'm glad to hear from different voices. πππ \nto prove that I just downvoted it and you can see it's 2 downvotesπππ
\n\nAnd to be nice and clear,is it strange that for a university project which funded by tax payers,to spend money to hold a contest which rejects obvious better solutions????
", "rawMarkdown": "I didn't downvote your reply. I welcome you to comment on my thread,and I'm glad to hear from different voices. πππ \nto prove that I just downvoted it and you can see it's 2 downvotesπππ\n\nAnd to be nice and clear,is it strange that for a university project which funded by tax payers,to spend money to hold a contest which rejects obvious better solutions????", "votes": -4 }, { "id": 923188, "postDate": "2020-07-10T15:41:47.247Z", "content": "@alexwang2019 Or, you could assume good intent on the part of Kaggle and the host, with both sides working diligently to ensure the best path forward given new details, and constantly trying to improve.
\nImagine how dysfunctional things would be here if everyone disobeyed rules they didn't agree with.
", "rawMarkdown": "@alexwang2019 Or, you could assume good intent on the part of Kaggle and the host, with both sides working diligently to ensure the best path forward given new details, and constantly trying to improve.\n\nImagine how dysfunctional things would be here if everyone disobeyed rules they didn't agree with.", "votes": 5 } ] }, { "id": 922606, "postDate": "2020-07-10T08:24:44.823Z", "content": "Couldn't all the notebooks that link to a yolo5 data source be invalidated as a starting point if it was to be enforced? You could make your own and change it to get around it but it would at least shakeup the leaderboard as a starting point.\nIt is a bit hard as it's more annoying to throw away work than to be glad you didn't start using yolov5, so it's going to be unfair to someone.
", "rawMarkdown": "Couldn't all the notebooks that link to a yolo5 data source be invalidated as a starting point if it was to be enforced? You could make your own and change it to get around it but it would at least shakeup the leaderboard as a starting point.\nIt is a bit hard as it's more annoying to throw away work than to be glad you didn't start using yolov5, so it's going to be unfair to someone." }, { "id": 922511, "postDate": "2020-07-10T06:50:55.763Z", "content": "If i implement yolov5 by myself, is it ok?
", "rawMarkdown": "If i implement yolov5 by myself, is it ok?", "replies": [ { "id": 922543, "postDate": "2020-07-10T07:28:02.433Z", "content": "It depends what you mean exactly by \"implement yolov5 yourself\". From what I understand (note: I haven't used yolov5), is that it isn't a novel network architecture that has been published in an article or anything. Rather, it is a software package that builds on the older yolov3.
\n\nSo when you go to implement yolov5, what are you basing it off of? The yolov3 paper or the yolov5 code itself? I would assume that copying idea for idea from the yolov5 source code could be considered copyright infringement but I'm certainly no expert. Whereas building your own implementation from the paper would not.
", "rawMarkdown": "It depends what you mean exactly by \"implement yolov5 yourself\". From what I understand (note: I haven't used yolov5), is that it isn't a novel network architecture that has been published in an article or anything. Rather, it is a software package that builds on the older yolov3.\n\nSo when you go to implement yolov5, what are you basing it off of? The yolov3 paper or the yolov5 code itself? I would assume that copying idea for idea from the yolov5 source code could be considered copyright infringement but I'm certainly no expert. Whereas building your own implementation from the paper would not." }, { "id": 922550, "postDate": "2020-07-10T07:33:29.033Z", "content": "yolov5 has nothing new, just bunch of tricks . I can name it as yolo6.\nAnd there are some disccusion about the name , \"yolo5\", already.
", "rawMarkdown": "yolov5 has nothing new, just bunch of tricks . I can name it as yolo6.\nAnd there are some disccusion about the name , \"yolo5\", already.\n\n", "votes": 2 }, { "id": 922574, "postDate": "2020-07-10T07:57:09.947Z", "content": "Yeah, that's what I mean. If you're just taking the \"tricks\" they implemented in Yolov5, it's probably not allowed.
", "rawMarkdown": "Yeah, that's what I mean. If you're just taking the \"tricks\" they implemented in Yolov5, it's probably not allowed." }, { "id": 922844, "postDate": "2020-07-10T11:08:41.123Z", "content": "what tricks is created by yolo5?
", "rawMarkdown": "what tricks is created by yolo5?" } ] }, { "id": 921496, "postDate": "2020-07-09T10:36:16.760Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 921430, "author_name": "Peiyuan Liao", "author_url": "", "post_date": "2020-07-09T09:14:22.170000", "content": "Also, I'm still confused about a point I brought up in another thread, which is as follows: say hypothetically one of the top three teams ended up using Yolov5 or other code not permitted by the competition. Do they move on to the next team up on the leaderboard to check for eligibility and straight up delete the non-compliant team? What if this goes on for a while and potentially lots of teams get disqualified because they (originally was only in say the gold medal zone) moved to the \"in-the-money\" zone (because the teams before them got deleted) and get checked for eligibility? I'm looking for a formal answer from the organizer because the answer to this affects which pipeline I choose to develop in the following weeks. Yolov5 indeed works well but there are other strong-performing models that I developed. Thanks!
", "votes": 9, "replies": [] }, { "id": 921413, "author_name": "MichaelP", "author_url": "", "post_date": "2020-07-09T08:56:14.227000", "content": "This is a good point, I've also been considering this. Possibly on this occasion Kaggle could check all gold medal winning entries not just the prize zone? Otherwise it feels pretty unfair to people who do stick to the clarified rules. Hopefully other solutions will prove to be equally effective as Etienne suggests. Also his comment makes it sound like a shakeup is on the way as private and public test sets are different.
", "votes": 3, "replies": [] }, { "id": 921270, "author_name": "Etienne David", "author_url": "", "post_date": "2020-07-09T07:08:05.887000", "content": "It is a possible situation, but there is still few days before the end of the competition. The private test set is also different from the public test set and will be corrected, it can also shake up the final leaderboard !
", "votes": 2, "replies": [] }, { "id": 921554, "author_name": "Ron Dudkman", "author_url": "", "post_date": "2020-07-09T11:27:38.043000", "content": "In case the 3 teams at \"in the money\" gets disqualifed, that means the 4th-6th teams will climb to 1st-3rd places?\nBecause if thats true, the 4th-6th teams will also gets disqualifed (probably), and if we continue on that way everyone at the top of LB will get disqualifed untill we find the first team who did not use Yolov5.
\n\nSo yeah there is a paradox here, because if I use yolov5 I want to get higher score as I can, but on the other hand I should hope to get lower score than the real winners (who did not use Yolov5) so I wont get disqualifed.
\n\nI think the best solution is to not disqualify anyone, and give the money prizes for the higest score who stand in the competition rules (which will probably be at 20+ at LB).
", "votes": 2, "replies": [ { "id": 921618, "author_name": "Peiyuan Liao", "author_url": "", "post_date": "2020-07-09T12:27:34.917000", "content": "Well, the problem for that is some of the competitors care more about the rank than the prize money, and if they complied with the rules from the beginning, that would be unfair for them.
", "votes": 4, "replies": [] }, { "id": 921695, "author_name": "MichaelP", "author_url": "", "post_date": "2020-07-09T13:37:43.753000", "content": "Yes I completely agree, for some people the rank and especially a gold medal is the main motivation.
", "votes": 2, "replies": [] } ] }, { "id": 922361, "author_name": "Alex_Wang", "author_url": "", "post_date": "2020-07-10T03:52:46.017000", "content": "The License rule is clarified just a few days ago, and right in the middle of the competition period time. \nTo me, it is quite unfair for all people participate here, because they just did their part, yolov5 or not, everyone must pick up something to begin, no one can compete from building a framework.\nSo, it is clear the host's fault and responsibility, they should anticipate this situation, and they can make this right!\nThe host need only make an enquiry to YOLOV5 owner to get a license, pay some money to get the highest performance.
", "votes": 1, "replies": [ { "id": 922535, "author_name": "Matt", "author_url": "", "post_date": "2020-07-10T07:23:46.927000", "content": "In my opinion, it wouldn't be fair to allow it at this point. That would amount to changing the rules which we all agreed to when joining the competition. It would be unfair to the people who abided by the licencing rule and have already made submissions that fit the rule. I agree it's not the best situation, but the best thing to do is learn our lesson and not repeat our mistakes in the future.
", "votes": 1, "replies": [] }, { "id": 922626, "author_name": "Alex_Wang", "author_url": "", "post_date": "2020-07-10T08:37:24.787000", "content": "The ridiculousness of this rule is like to ban BERT for an NLP competition, rule out pretrained weights for any CV competition, not allowing Pi doing a competition on quadrature a circle .rule out black people from doing something
\n\nNot every rule is a good rule ,thinking of Hitler's Nurnberg Laws. sometimes, it need speople to disobey the rules to make things right!!!!
", "votes": -6, "replies": [] }, { "id": 922700, "author_name": "Matt", "author_url": "", "post_date": "2020-07-10T09:25:33.280000", "content": "It's okay to disagree with my opinion and downvote my comment as you have. But comparing the Kaggle rules to racism and the Nuremberg laws is a little too far. Just remember the Kaggle etiquette of \"Be patient, be friendly, and focus on ideas\".
", "votes": 6, "replies": [] }, { "id": 922916, "author_name": "Alex_Wang", "author_url": "", "post_date": "2020-07-10T12:16:55.510000", "content": "I didn't downvote your reply. I welcome you to comment on my thread,and I'm glad to hear from different voices. πππ \nto prove that I just downvoted it and you can see it's 2 downvotesπππ
\n\nAnd to be nice and clear,is it strange that for a university project which funded by tax payers,to spend money to hold a contest which rejects obvious better solutions????
", "votes": -4, "replies": [] }, { "id": 923188, "author_name": "inversion", "author_url": "", "post_date": "2020-07-10T15:41:47.247000", "content": "@alexwang2019 Or, you could assume good intent on the part of Kaggle and the host, with both sides working diligently to ensure the best path forward given new details, and constantly trying to improve.
\nImagine how dysfunctional things would be here if everyone disobeyed rules they didn't agree with.
", "votes": 5, "replies": [] } ] }, { "id": 922606, "author_name": "peachthief", "author_url": "", "post_date": "2020-07-10T08:24:44.823000", "content": "Couldn't all the notebooks that link to a yolo5 data source be invalidated as a starting point if it was to be enforced? You could make your own and change it to get around it but it would at least shakeup the leaderboard as a starting point.\nIt is a bit hard as it's more annoying to throw away work than to be glad you didn't start using yolov5, so it's going to be unfair to someone.
", "votes": 0, "replies": [] }, { "id": 922511, "author_name": "coolz", "author_url": "", "post_date": "2020-07-10T06:50:55.763000", "content": "If i implement yolov5 by myself, is it ok?
", "votes": 0, "replies": [ { "id": 922543, "author_name": "Matt", "author_url": "", "post_date": "2020-07-10T07:28:02.433000", "content": "It depends what you mean exactly by \"implement yolov5 yourself\". From what I understand (note: I haven't used yolov5), is that it isn't a novel network architecture that has been published in an article or anything. Rather, it is a software package that builds on the older yolov3.
\n\nSo when you go to implement yolov5, what are you basing it off of? The yolov3 paper or the yolov5 code itself? I would assume that copying idea for idea from the yolov5 source code could be considered copyright infringement but I'm certainly no expert. Whereas building your own implementation from the paper would not.
", "votes": 0, "replies": [] }, { "id": 922550, "author_name": "coolz", "author_url": "", "post_date": "2020-07-10T07:33:29.033000", "content": "yolov5 has nothing new, just bunch of tricks . I can name it as yolo6.\nAnd there are some disccusion about the name , \"yolo5\", already.
", "votes": 2, "replies": [] }, { "id": 922574, "author_name": "Matt", "author_url": "", "post_date": "2020-07-10T07:57:09.947000", "content": "Yeah, that's what I mean. If you're just taking the \"tricks\" they implemented in Yolov5, it's probably not allowed.
", "votes": 0, "replies": [] }, { "id": 922844, "author_name": "coolz", "author_url": "", "post_date": "2020-07-10T11:08:41.123000", "content": "what tricks is created by yolo5?
", "votes": 0, "replies": [] } ] }, { "id": 921496, "author_name": "", "author_url": "", "post_date": "2020-07-09T10:36:16.760000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "921264": "As all YoloV5 solutions are now invalidated because of the incompatible license. But as of now, they are working best (based on the leaderboard, ~almost all top 50 teams use it) **so far**. It's also mentioned by competition host that only winning solutions (top 3) have to be MIT licensed but the license of solutions in the medal zone doesn't matter much.\n\nNow, having said that (and assuming **Yolov5 will be the best till the end**), here is the paradox: To win the competition, you will have to use Yolov5 (which will be invalidated if you actually win). If you are using techniques other than Yolov5 but that doesn't score as high as Yolov5, you won't be in the top three in the first place (even if your solution is second best after Yolov5 and eligible to win) because of Yolov5 submissions by the guys who want to be in medal zone (actually everyone who uses Yolov5, because who doesn't want medals, right?) will be there ahead of you in the leaderboard.\n\nThere are some assumptions in this (like yolov5 is best and no other MIT licensed approach gives scores as high as Yolov5), but they are valid as per the current situation of the competition.\n\n@maggiemd , @bendvd What do you think about this?", "921430": "Also, I'm still confused about a point I brought up in another thread, which is as follows: say hypothetically one of the top three teams ended up using Yolov5 or other code not permitted by the competition. Do they move on to the next team up on the leaderboard to check for eligibility and straight up delete the non-compliant team? What if this goes on for a while and potentially lots of teams get disqualified because they (originally was only in say the gold medal zone) moved to the \"in-the-money\" zone (because the teams before them got deleted) and get checked for eligibility? I'm looking for a formal answer from the organizer because the answer to this affects which pipeline I choose to develop in the following weeks. Yolov5 indeed works well but there are other strong-performing models that I developed. Thanks!", "921413": "This is a good point, I've also been considering this. Possibly on this occasion Kaggle could check all gold medal winning entries not just the prize zone? Otherwise it feels pretty unfair to people who do stick to the clarified rules. Hopefully other solutions will prove to be equally effective as Etienne suggests. Also his comment makes it sound like a shakeup is on the way as private and public test sets are different.", "921270": "It is a possible situation, but there is still few days before the end of the competition. The private test set is also different from the public test set and will be corrected, it can also shake up the final leaderboard !", "921554": "In case the 3 teams at \"in the money\" gets disqualifed, that means the 4th-6th teams will climb to 1st-3rd places?\nBecause if thats true, the 4th-6th teams will also gets disqualifed (probably), and if we continue on that way everyone at the top of LB will get disqualifed untill we find the first team who did not use Yolov5.\n\nSo yeah there is a paradox here, because if I use yolov5 I want to get higher score as I can, but on the other hand I should hope to get lower score than the real winners (who did not use Yolov5) so I wont get disqualifed.\n\nI think the best solution is to not disqualify anyone, and give the money prizes for the higest score who stand in the competition rules (which will probably be at 20+ at LB).", "922361": "The License rule is clarified just a few days ago, and right in the middle of the competition period time. \nTo me, it is quite unfair for all people participate here, because they just did their part, yolov5 or not, everyone must pick up something to begin, no one can compete from building a framework.\nSo, it is clear the host's fault and responsibility, they should anticipate this situation, and they can make this right!\nThe host need only make an enquiry to YOLOV5 owner to get a license, pay some money to get the highest performance. ", "922606": "Couldn't all the notebooks that link to a yolo5 data source be invalidated as a starting point if it was to be enforced? You could make your own and change it to get around it but it would at least shakeup the leaderboard as a starting point.\nIt is a bit hard as it's more annoying to throw away work than to be glad you didn't start using yolov5, so it's going to be unfair to someone.", "922511": "If i implement yolov5 by myself, is it ok?", "921496": "" } }