![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2694078%2Ff44c7181031fb0f88aead591d428e5c8%2FTIM1.jpg?generation=1574138137431374&alt=media\")
Earlier today：\n
But now：\nAll my submissions and recordings are disappear！\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2694078%2Fb374663155dc3e20478f7980004594b2%2FTIM2.png?generation=1574138273581485&alt=media\")
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2694078%2Fc4c32047e411c89beebfd92fa7e68bfc%2FTIM2.png?generation=1574138296855639&alt=media\")
Earlier today：\n
But now：\nAll my submissions and recordings are disappear！\n\n
I want to know how long does it take to get the result of the appeal?
", "rawMarkdown": "I want to know how long does it take to get the result of the appeal?", "votes": 3, "replies": [ { "id": 676650, "postDate": "2019-11-19T11:45:41.930Z", "content": "我已经得到回复了，可能时区关系他们还在休息，估计明早就能有结果了，我的情况是这样的：我们实验室本来一队人想一起参加的，可是由于过了那个合并的截止日期，然后一个人没合上，单独一个队伍，我们给他分享了提交文件，他也提交了这个结果，他们检测系统估计是根据提交文件来的，然后就把我们几个有相似提交记录的都封了，不知道你有没有这种情况，我们几个本来这场之后应该都是expert了吧，想想真的蛮可惜的
", "rawMarkdown": "我已经得到回复了，可能时区关系他们还在休息，估计明早就能有结果了，我的情况是这样的：我们实验室本来一队人想一起参加的，可是由于过了那个合并的截止日期，然后一个人没合上，单独一个队伍，我们给他分享了提交文件，他也提交了这个结果，他们检测系统估计是根据提交文件来的，然后就把我们几个有相似提交记录的都封了，不知道你有没有这种情况，我们几个本来这场之后应该都是expert了吧，想想真的蛮可惜的" }, { "id": 676722, "postDate": "2019-11-19T13:10:56.797Z", "content": "kaggle通过邮箱答复？请问你是大概什么时间提交的申诉呢？ 我们大约14:00（北京时间）提交的。\n我们没有这种情况，我们前一阶段只有四个人的账号在提交，最后20天左右才加入了第五位同学。\n我们也没有与其他队伍分享过任何代码和文件，很神奇。
", "rawMarkdown": "kaggle通过邮箱答复？请问你是大概什么时间提交的申诉呢？ 我们大约14:00（北京时间）提交的。\n我们没有这种情况，我们前一阶段只有四个人的账号在提交，最后20天左右才加入了第五位同学。\n我们也没有与其他队伍分享过任何代码和文件，很神奇。" }, { "id": 676750, "postDate": "2019-11-19T13:32:56.783Z", "content": "@sj626591833 我觉得这种情况算违规的，因为他没在你们队伍里而你们分享给他结果了，申诉希望渺茫。挺可惜的，以后要注意了。。。
\n\nYour situation may still be regarded as cheating because your friend was not in your team when you shared results with him.
", "rawMarkdown": "@sj626591833 我觉得这种情况算违规的，因为他没在你们队伍里而你们分享给他结果了，申诉希望渺茫。挺可惜的，以后要注意了。。。\n\nYour situation may still be regarded as cheating because your friend was not in your team when you shared results with him." }, { "id": 676760, "postDate": "2019-11-19T13:39:55.780Z", "content": "@gogo827jz Thank you for your attention to our situation.\nWe have read the rules carefully, and our understanding is that n people can share code and data as long as they are eventually merged into the same team. We did according to this rule in APTOS(normal) and this competition.
", "rawMarkdown": "@gogo827jz Thank you for your attention to our situation.\nWe have read the rules carefully, and our understanding is that n people can share code and data as long as they are eventually merged into the same team. We did according to this rule in APTOS(normal) and this competition." }, { "id": 676779, "postDate": "2019-11-19T13:51:22.413Z", "content": "You can read this reply by Kaggle admin on the same issue just a few days back: https://www.kaggle.com/c/ashrae-energy-prediction/discussion/116260#670739
", "rawMarkdown": "You can read this reply by Kaggle admin on the same issue just a few days back: https://www.kaggle.com/c/ashrae-energy-prediction/discussion/116260#670739", "votes": 2 }, { "id": 676799, "postDate": "2019-11-19T14:10:11.670Z", "content": "Oh, thank you. This issue seems to be controversial. If kaggle's rule is that the five members in our team can't share code and data before we are eventually merged into the same team, I know why our team disappear.\nBut after reading the rules carefully, I still think we can share code and data before we are eventually merged into the same team. We did as this in APTOS and we got normal rank.\nWe are confused now.
", "rawMarkdown": "Oh, thank you. This issue seems to be controversial. If kaggle's rule is that the five members in our team can't share code and data before we are eventually merged into the same team, I know why our team disappear.\nBut after reading the rules carefully, I still think we can share code and data before we are eventually merged into the same team. We did as this in APTOS and we got normal rank.\nWe are confused now." }, { "id": 676821, "postDate": "2019-11-19T14:25:29.770Z", "rawMarkdown": "", "isDeleted": true }, { "id": 676836, "postDate": "2019-11-19T14:39:05.930Z", "content": "Yes, as you said, in several single days we make 25 submissions. Because we didn't start work right away when the competition started. We think it is fair that all teams have the same maximum number of submissions. If we did not merge into a team finally, or we exceeded the maximum number of submissions, we accept punishment.\nIf the way we did violate the kaggle's rules, we can only accept punishment and gain experience for future competitions.
", "rawMarkdown": "Yes, as you said, in several single days we make 25 submissions. Because we didn't start work right away when the competition started. We think it is fair that all teams have the same maximum number of submissions. If we did not merge into a team finally, or we exceeded the maximum number of submissions, we accept punishment.\nIf the way we did violate the kaggle's rules, we can only accept punishment and gain experience for future competitions.", "votes": 1 }, { "id": 677293, "postDate": "2019-11-20T02:17:40.973Z", "content": "是的遗憾归遗憾，主要自己操作失误了，不应该给其他队伍，主要是过了merge的时间，然后又出于好心，就把结果给了她
", "rawMarkdown": "是的遗憾归遗憾，主要自己操作失误了，不应该给其他队伍，主要是过了merge的时间，然后又出于好心，就把结果给了她" } ] }, { "id": 676461, "postDate": "2019-11-19T07:54:16.513Z", "content": "This is what I wrote when the private leaderboard was open. Even if we temporarily disappear on the leaderboard. I want to share our thought and our 151st solution.\nOur final submission's private score is 0.65553(141st now). Our top score is 0.65948(just luck). Our most stable submission's score is 0.65832(92nd now), but unfortunately we did not submit this version, we are too eager to score high on the public set.
\n\n# Some words:\nIt's my second competition in kaggle and once again I won a bronze medal(Almost dropped out of the medal area). I dropped 107 on the private dataset than the public dataset which is in my expectation.\nAfter my first competition(APTOS) I questioned the division of public and private dataset. But I gradually thought that the meaning of the game is not over-fitting the test set. Mechanically adjusting various thresholds does not produce particularly good results.\nBut this time, I chose the top public score one to submit once again, instead of choosing the most stable one without adjusting thresholds(a 4*TTA + 10fold version in the pic below, which rank 92nd on the private leaderboard). \n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2694078%2Fb89a09cbb23f263acd2d00d6180ec6be%2Fpic.png?generation=1574133598717370&alt=media\")
# My solution(most stable submission scoring 0.65832)：
\n\nAdd masks from stage1 and masks in keras-efficientnetb2-for-classifying-cloud together. Than erode them by 3x3 kernel once.
\n\nUse keras-efficientnetb2-for-classifying-cloud to del masks with low confidence.(I trained another better model than the open version.)
", "rawMarkdown": "This is what I wrote when the private leaderboard was open. Even if we temporarily disappear on the leaderboard. I want to share our thought and our 151st solution.\nOur final submission's private score is 0.65553(141st now). Our top score is 0.65948(just luck). Our most stable submission's score is 0.65832(92nd now), but unfortunately we did not submit this version, we are too eager to score high on the public set.\n\n**# Some words:**\nIt's my second competition in kaggle and once again I won a bronze medal(Almost dropped out of the medal area). I dropped 107 on the private dataset than the public dataset which is in my expectation.\nAfter my first competition(APTOS) I questioned the division of public and private dataset. But I gradually thought that the meaning of the game is not over-fitting the test set. Mechanically adjusting various thresholds does not produce particularly good results.\nBut this time, I chose the top public score one to submit once again, instead of choosing the most stable one without adjusting thresholds(a 4*TTA + 10fold version in the pic below, which rank 92nd on the private leaderboard). \n\n\n**# My solution(most stable submission scoring 0.65832)：**\n### stage1:\n- Network:UNet-SE_ResNext101\n- Augmentations: H/V flip, ShiftScalerRotate.\n- 10-fold\n- TTA: raw, Horizontal Flip, Vertical Flip, Horizontal+Vertical Flip\n\n### stage2:\nAdd masks from stage1 and masks in [keras-efficientnetb2-for-classifying-cloud](https://www.kaggle.com/mobassir/keras-efficientnetb2-for-classifying-cloud) together. Than erode them by 3x3 kernel once.\n\n### stage3:\nUse [keras-efficientnetb2-for-classifying-cloud](https://www.kaggle.com/mobassir/keras-efficientnetb2-for-classifying-cloud) to del masks with low confidence.(I trained another better model than the open version.)", "votes": 1 }, { "id": 676412, "postDate": "2019-11-19T06:35:31.660Z", "content": "I have contacted compliance and submit our explanation. I want to know why our team disappear, so I can provide more targeted evidence.\nThank you @inversion . And can you tell me what can we do next?
", "rawMarkdown": "I have contacted compliance and submit our explanation. I want to know why our team disappear, so I can provide more targeted evidence.\nThank you @inversion . And can you tell me what can we do next?", "votes": 1 }, { "id": 677730, "postDate": "2019-11-20T14:30:34.397Z", "content": "![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2694078%2F24ecc3256d2e7e9c249c5882dcd09cf5%2Ftim4.png?generation=1574259796980982&alt=media\")
We finally received an official response from kaggle, we have misinterpreted the rules of the game. It is true that we shared code and data before we were eventually merged into the same team. What a pity! As a newcomer to kaggle(it is our second competition), what we can do is to accept and learn from the experience.
\n\nThank you all for your attention and help. See you next game.
", "rawMarkdown": "\n\nWe finally received an official response from kaggle, we have misinterpreted the rules of the game. It is true that we shared code and data before we were eventually merged into the same team. What a pity! As a newcomer to kaggle(it is our second competition), what we can do is to accept and learn from the experience.\n\nThank you all for your attention and help. See you next game.", "votes": 2, "replies": [ { "id": 677752, "postDate": "2019-11-20T14:53:57.383Z", "content": "no offence, it was not pity, I think.. even you are a newcomer, you should know that the behavior of sharing code with different team is not allowed, included you can not share any methods about the competition----called private sharing\nI believe that all the upright kagglers hate this kind of behavior。\nbtw, I hope you can keep this lesson in mind and get great score next time, good luck.
", "rawMarkdown": "no offence, it was not pity, I think.. even you are a newcomer, you should know that the behavior of sharing code with different team is not allowed, included you can not share any methods about the competition----called **private sharing**\nI believe that all the upright kagglers hate this kind of behavior。\nbtw, I hope you can keep this lesson in mind and get great score next time, good luck.\n\n" }, { "id": 678083, "postDate": "2019-11-21T01:25:38.563Z", "content": "Yeal, maybe English is not our native language or we had an error understanding of this rule:\n“This prohibition includes sharing Competition Code between separate Teams, unless a Team merger occurs.”
\nWe worked together before we merge, this is a fact, we recognize.
It is a lesson from kaggle for us. Thank you.
", "rawMarkdown": "Yeal, maybe English is not our native language or we had an error understanding of this rule:\n“This prohibition includes sharing Competition Code between separate Teams, unless a Team merger occurs.” \nWe worked together before we merge, this is a fact, we recognize.\n\nIt is a lesson from kaggle for us. Thank you.", "votes": 1 } ] }, { "id": 676642, "postDate": "2019-11-19T11:40:43.367Z", "content": "As a member of our team, I am waiting for the appeal result.
", "rawMarkdown": "As a member of our team, I am waiting for the appeal result." }, { "id": 676305, "postDate": "2019-11-19T04:41:14.230Z", "content": "@lifengnan I believe someone from your team used multiple accounts.
", "rawMarkdown": "@lifengnan I believe someone from your team used multiple accounts. ", "replies": [ { "id": 676309, "postDate": "2019-11-19T04:46:04.917Z", "content": "Oh, our team has five people, one person one account. It just because we have use the same computer to submit when we were on a business trip? I can't believe it!
", "rawMarkdown": "Oh, our team has five people, one person one account. It just because we have use the same computer to submit when we were on a business trip? I can't believe it!" }, { "id": 676315, "postDate": "2019-11-19T04:51:18.550Z", "content": "You can submit a ticket to kaggle and explain your situation. If you are innocent, they may consider recovering for you.
", "rawMarkdown": "You can submit a ticket to kaggle and explain your situation. If you are innocent, they may consider recovering for you.", "votes": 2 }, { "id": 676319, "postDate": "2019-11-19T04:53:43.730Z", "content": "Yeah, Yirun is right kaggle might consider your situation.
", "rawMarkdown": "Yeah, Yirun is right kaggle might consider your situation." }, { "id": 676323, "postDate": "2019-11-19T04:59:17.683Z", "content": "Thanks. Do you mean explain in this way?\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2694078%2Faef943bbd14303a8ba11849893dcddc0%2FTIM3.png?generation=1574139509064474&alt=media\")
Yes
", "rawMarkdown": "Yes" }, { "id": 676327, "postDate": "2019-11-19T05:03:40.207Z", "content": "You should definitely contact compliance.
\n\nhttps://www.kaggle.com/compliance
\n\nOut of curiosity, why would you think there would be any problem if 5 different team members all submitted from the same computer? (It's not a problem, as long as you're not sharing outside of your team.)
", "rawMarkdown": "You should definitely contact compliance.\n\nhttps://www.kaggle.com/compliance\n\nOut of curiosity, why would you think there would be any problem if 5 different team members all submitted from the same computer? (It's not a problem, as long as you're not sharing outside of your team.)" }, { "id": 676328, "postDate": "2019-11-19T05:05:39.287Z", "content": "Maybe because of IP address validation for each submission?
", "rawMarkdown": "Maybe because of IP address validation for each submission?" }, { "id": 676330, "postDate": "2019-11-19T05:09:07.510Z", "content": "What I mean is, as long as everyone is on the same team, why would submitting from the same IP or the same computer be suspicious? I certainly wouldn't think it's suspicious.
", "rawMarkdown": "What I mean is, as long as everyone is on the same team, why would submitting from the same IP or the same computer be suspicious? I certainly wouldn't think it's suspicious." }, { "id": 676332, "postDate": "2019-11-19T05:10:05.627Z", "content": "Anyway, compliance is good at digging into these things and will get things sorted out appropriately.
", "rawMarkdown": "Anyway, compliance is good at digging into these things and will get things sorted out appropriately." }, { "id": 676341, "postDate": "2019-11-19T05:16:13.380Z", "content": "Thanks. But I can't open this link, it may have some problem.\nWhat can we do? Is there any other way of explaining?\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2694078%2Fe72bf2fcf8dad7087f6aa715e841e2a1%2FTIM5.png?generation=1574140186870444&alt=media\")
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2694078%2Ff68bea37a8e98661601ed34cc8bd9b9d%2FTI4.png?generation=1574140186870794&alt=media\")
Because it is the only possibility I can think of. We five are classmates in a lab.
", "rawMarkdown": "Because it is the only possibility I can think of. We five are classmates in a lab." }, { "id": 676358, "postDate": "2019-11-19T05:37:59.043Z", "content": "Thanks, I opened this link successfully.
", "rawMarkdown": "Thanks, I opened this link successfully." }, { "id": 676361, "postDate": "2019-11-19T05:45:08.197Z", "content": "I came across the same situation here,I am so depressed to find our team disappearing on the leaderboard, I mean we guys are actually in the same lab,and due to the blocking technique of chinese great wall ,so we submitted our results here with same showsocks IP address,if the official needs any proof on our work in the past month.I would like to upload everything we've got here.
", "rawMarkdown": "I came across the same situation here,I am so depressed to find our team disappearing on the leaderboard, I mean we guys are actually in the same lab,and due to the blocking technique of chinese great wall ,so we submitted our results here with same showsocks IP address,if the official needs any proof on our work in the past month.I would like to upload everything we've got here.", "votes": 1 }, { "id": 676365, "postDate": "2019-11-19T05:48:20.387Z", "content": "Yeal, we are at Beijing,China. The same situation.
", "rawMarkdown": "Yeal, we are at Beijing,China. The same situation.", "votes": 1 }, { "id": 676367, "postDate": "2019-11-19T05:48:57.737Z", "content": "Yes, in China, the submit process needs VPN...
", "rawMarkdown": "Yes, in China, the submit process needs VPN...", "votes": 1 } ] } ], "comments": [ { "id": 676565, "author_name": "lifengnan", "author_url": "", "post_date": "2019-11-19T10:02:56.670000", "content": "I want to know how long does it take to get the result of the appeal?
", "votes": 3, "replies": [ { "id": 676650, "author_name": "Overfit Queen", "author_url": "", "post_date": "2019-11-19T11:45:41.930000", "content": "我已经得到回复了，可能时区关系他们还在休息，估计明早就能有结果了，我的情况是这样的：我们实验室本来一队人想一起参加的，可是由于过了那个合并的截止日期，然后一个人没合上，单独一个队伍，我们给他分享了提交文件，他也提交了这个结果，他们检测系统估计是根据提交文件来的，然后就把我们几个有相似提交记录的都封了，不知道你有没有这种情况，我们几个本来这场之后应该都是expert了吧，想想真的蛮可惜的
", "votes": 0, "replies": [] }, { "id": 676722, "author_name": "lifengnan", "author_url": "", "post_date": "2019-11-19T13:10:56.797000", "content": "kaggle通过邮箱答复？请问你是大概什么时间提交的申诉呢？ 我们大约14:00（北京时间）提交的。\n我们没有这种情况，我们前一阶段只有四个人的账号在提交，最后20天左右才加入了第五位同学。\n我们也没有与其他队伍分享过任何代码和文件，很神奇。
", "votes": 0, "replies": [] }, { "id": 676750, "author_name": "Yirun Zhang", "author_url": "", "post_date": "2019-11-19T13:32:56.783000", "content": "@sj626591833 我觉得这种情况算违规的，因为他没在你们队伍里而你们分享给他结果了，申诉希望渺茫。挺可惜的，以后要注意了。。。
\n\nYour situation may still be regarded as cheating because your friend was not in your team when you shared results with him.
", "votes": 0, "replies": [] }, { "id": 676760, "author_name": "lifengnan", "author_url": "", "post_date": "2019-11-19T13:39:55.780000", "content": "@gogo827jz Thank you for your attention to our situation.\nWe have read the rules carefully, and our understanding is that n people can share code and data as long as they are eventually merged into the same team. We did according to this rule in APTOS(normal) and this competition.
", "votes": 0, "replies": [] }, { "id": 676779, "author_name": "Vopani", "author_url": "", "post_date": "2019-11-19T13:51:22.413000", "content": "You can read this reply by Kaggle admin on the same issue just a few days back: https://www.kaggle.com/c/ashrae-energy-prediction/discussion/116260#670739
", "votes": 2, "replies": [] }, { "id": 676799, "author_name": "lifengnan", "author_url": "", "post_date": "2019-11-19T14:10:11.670000", "content": "Oh, thank you. This issue seems to be controversial. If kaggle's rule is that the five members in our team can't share code and data before we are eventually merged into the same team, I know why our team disappear.\nBut after reading the rules carefully, I still think we can share code and data before we are eventually merged into the same team. We did as this in APTOS and we got normal rank.\nWe are confused now.
", "votes": 0, "replies": [] }, { "id": 676821, "author_name": "", "author_url": "", "post_date": "2019-11-19T14:25:29.770000", "content": "", "votes": 0, "replies": [] }, { "id": 676836, "author_name": "lifengnan", "author_url": "", "post_date": "2019-11-19T14:39:05.930000", "content": "Yes, as you said, in several single days we make 25 submissions. Because we didn't start work right away when the competition started. We think it is fair that all teams have the same maximum number of submissions. If we did not merge into a team finally, or we exceeded the maximum number of submissions, we accept punishment.\nIf the way we did violate the kaggle's rules, we can only accept punishment and gain experience for future competitions.
", "votes": 1, "replies": [] }, { "id": 677293, "author_name": "Overfit Queen", "author_url": "", "post_date": "2019-11-20T02:17:40.973000", "content": "是的遗憾归遗憾，主要自己操作失误了，不应该给其他队伍，主要是过了merge的时间，然后又出于好心，就把结果给了她
", "votes": 0, "replies": [] } ] }, { "id": 676461, "author_name": "lifengnan", "author_url": "", "post_date": "2019-11-19T07:54:16.513000", "content": "This is what I wrote when the private leaderboard was open. Even if we temporarily disappear on the leaderboard. I want to share our thought and our 151st solution.\nOur final submission's private score is 0.65553(141st now). Our top score is 0.65948(just luck). Our most stable submission's score is 0.65832(92nd now), but unfortunately we did not submit this version, we are too eager to score high on the public set.
\n\n# Some words:\nIt's my second competition in kaggle and once again I won a bronze medal(Almost dropped out of the medal area). I dropped 107 on the private dataset than the public dataset which is in my expectation.\nAfter my first competition(APTOS) I questioned the division of public and private dataset. But I gradually thought that the meaning of the game is not over-fitting the test set. Mechanically adjusting various thresholds does not produce particularly good results.\nBut this time, I chose the top public score one to submit once again, instead of choosing the most stable one without adjusting thresholds(a 4*TTA + 10fold version in the pic below, which rank 92nd on the private leaderboard). \n
# My solution(most stable submission scoring 0.65832)：
\n\nAdd masks from stage1 and masks in keras-efficientnetb2-for-classifying-cloud together. Than erode them by 3x3 kernel once.
\n\nUse keras-efficientnetb2-for-classifying-cloud to del masks with low confidence.(I trained another better model than the open version.)
", "votes": 1, "replies": [] }, { "id": 676412, "author_name": "lifengnan", "author_url": "", "post_date": "2019-11-19T06:35:31.660000", "content": "I have contacted compliance and submit our explanation. I want to know why our team disappear, so I can provide more targeted evidence.\nThank you @inversion . And can you tell me what can we do next?
", "votes": 1, "replies": [] }, { "id": 677730, "author_name": "lifengnan", "author_url": "", "post_date": "2019-11-20T14:30:34.397000", "content": "
We finally received an official response from kaggle, we have misinterpreted the rules of the game. It is true that we shared code and data before we were eventually merged into the same team. What a pity! As a newcomer to kaggle(it is our second competition), what we can do is to accept and learn from the experience.
\n\nThank you all for your attention and help. See you next game.
", "votes": 2, "replies": [ { "id": 677752, "author_name": "Gary", "author_url": "", "post_date": "2019-11-20T14:53:57.383000", "content": "no offence, it was not pity, I think.. even you are a newcomer, you should know that the behavior of sharing code with different team is not allowed, included you can not share any methods about the competition----called private sharing\nI believe that all the upright kagglers hate this kind of behavior。\nbtw, I hope you can keep this lesson in mind and get great score next time, good luck.
", "votes": 0, "replies": [] }, { "id": 678083, "author_name": "lifengnan", "author_url": "", "post_date": "2019-11-21T01:25:38.563000", "content": "Yeal, maybe English is not our native language or we had an error understanding of this rule:\n“This prohibition includes sharing Competition Code between separate Teams, unless a Team merger occurs.”
\nWe worked together before we merge, this is a fact, we recognize.
It is a lesson from kaggle for us. Thank you.
", "votes": 1, "replies": [] } ] }, { "id": 676642, "author_name": "YongLi", "author_url": "", "post_date": "2019-11-19T11:40:43.367000", "content": "As a member of our team, I am waiting for the appeal result.
", "votes": 0, "replies": [] }, { "id": 676305, "author_name": "Ram Ramrakhya", "author_url": "", "post_date": "2019-11-19T04:41:14.230000", "content": "@lifengnan I believe someone from your team used multiple accounts.
", "votes": 0, "replies": [ { "id": 676309, "author_name": "lifengnan", "author_url": "", "post_date": "2019-11-19T04:46:04.917000", "content": "Oh, our team has five people, one person one account. It just because we have use the same computer to submit when we were on a business trip? I can't believe it!
", "votes": 0, "replies": [] }, { "id": 676315, "author_name": "Yirun Zhang", "author_url": "", "post_date": "2019-11-19T04:51:18.550000", "content": "You can submit a ticket to kaggle and explain your situation. If you are innocent, they may consider recovering for you.
", "votes": 2, "replies": [] }, { "id": 676319, "author_name": "Ram Ramrakhya", "author_url": "", "post_date": "2019-11-19T04:53:43.730000", "content": "Yeah, Yirun is right kaggle might consider your situation.
", "votes": 0, "replies": [] }, { "id": 676323, "author_name": "lifengnan", "author_url": "", "post_date": "2019-11-19T04:59:17.683000", "content": "Thanks. Do you mean explain in this way?\n
Yes
", "votes": 0, "replies": [] }, { "id": 676327, "author_name": "inversion", "author_url": "", "post_date": "2019-11-19T05:03:40.207000", "content": "You should definitely contact compliance.
\n\nhttps://www.kaggle.com/compliance
\n\nOut of curiosity, why would you think there would be any problem if 5 different team members all submitted from the same computer? (It's not a problem, as long as you're not sharing outside of your team.)
", "votes": 0, "replies": [] }, { "id": 676328, "author_name": "Ram Ramrakhya", "author_url": "", "post_date": "2019-11-19T05:05:39.287000", "content": "Maybe because of IP address validation for each submission?
", "votes": 0, "replies": [] }, { "id": 676330, "author_name": "inversion", "author_url": "", "post_date": "2019-11-19T05:09:07.510000", "content": "What I mean is, as long as everyone is on the same team, why would submitting from the same IP or the same computer be suspicious? I certainly wouldn't think it's suspicious.
", "votes": 0, "replies": [] }, { "id": 676332, "author_name": "inversion", "author_url": "", "post_date": "2019-11-19T05:10:05.627000", "content": "Anyway, compliance is good at digging into these things and will get things sorted out appropriately.
", "votes": 0, "replies": [] }, { "id": 676341, "author_name": "lifengnan", "author_url": "", "post_date": "2019-11-19T05:16:13.380000", "content": "Thanks. But I can't open this link, it may have some problem.\nWhat can we do? Is there any other way of explaining?\n\n
Because it is the only possibility I can think of. We five are classmates in a lab.
", "votes": 0, "replies": [] }, { "id": 676358, "author_name": "lifengnan", "author_url": "", "post_date": "2019-11-19T05:37:59.043000", "content": "Thanks, I opened this link successfully.
", "votes": 0, "replies": [] }, { "id": 676361, "author_name": "Overfit Queen", "author_url": "", "post_date": "2019-11-19T05:45:08.197000", "content": "I came across the same situation here,I am so depressed to find our team disappearing on the leaderboard, I mean we guys are actually in the same lab,and due to the blocking technique of chinese great wall ,so we submitted our results here with same showsocks IP address,if the official needs any proof on our work in the past month.I would like to upload everything we've got here.
", "votes": 1, "replies": [] }, { "id": 676365, "author_name": "lifengnan", "author_url": "", "post_date": "2019-11-19T05:48:20.387000", "content": "Yeal, we are at Beijing,China. The same situation.
", "votes": 1, "replies": [] }, { "id": 676367, "author_name": "Endi Niu", "author_url": "", "post_date": "2019-11-19T05:48:57.737000", "content": "Yes, in China, the submit process needs VPN...
", "votes": 1, "replies": [] } ] } ], "raw_markdown_by_id": { "676303": "Earlier today：\n\n\nBut now：\nAll my submissions and recordings are disappear！\n\n\n", "676565": "I want to know how long does it take to get the result of the appeal?", "676461": "This is what I wrote when the private leaderboard was open. Even if we temporarily disappear on the leaderboard. I want to share our thought and our 151st solution.\nOur final submission's private score is 0.65553(141st now). Our top score is 0.65948(just luck). Our most stable submission's score is 0.65832(92nd now), but unfortunately we did not submit this version, we are too eager to score high on the public set.\n\n**# Some words:**\nIt's my second competition in kaggle and once again I won a bronze medal(Almost dropped out of the medal area). I dropped 107 on the private dataset than the public dataset which is in my expectation.\nAfter my first competition(APTOS) I questioned the division of public and private dataset. But I gradually thought that the meaning of the game is not over-fitting the test set. Mechanically adjusting various thresholds does not produce particularly good results.\nBut this time, I chose the top public score one to submit once again, instead of choosing the most stable one without adjusting thresholds(a 4*TTA + 10fold version in the pic below, which rank 92nd on the private leaderboard). \n\n\n**# My solution(most stable submission scoring 0.65832)：**\n### stage1:\n- Network:UNet-SE_ResNext101\n- Augmentations: H/V flip, ShiftScalerRotate.\n- 10-fold\n- TTA: raw, Horizontal Flip, Vertical Flip, Horizontal+Vertical Flip\n\n### stage2:\nAdd masks from stage1 and masks in [keras-efficientnetb2-for-classifying-cloud](https://www.kaggle.com/mobassir/keras-efficientnetb2-for-classifying-cloud) together. Than erode them by 3x3 kernel once.\n\n### stage3:\nUse [keras-efficientnetb2-for-classifying-cloud](https://www.kaggle.com/mobassir/keras-efficientnetb2-for-classifying-cloud) to del masks with low confidence.(I trained another better model than the open version.)", "676412": "I have contacted compliance and submit our explanation. I want to know why our team disappear, so I can provide more targeted evidence.\nThank you @inversion . And can you tell me what can we do next?", "677730": "\n\nWe finally received an official response from kaggle, we have misinterpreted the rules of the game. It is true that we shared code and data before we were eventually merged into the same team. What a pity! As a newcomer to kaggle(it is our second competition), what we can do is to accept and learn from the experience.\n\nThank you all for your attention and help. See you next game.", "676642": "As a member of our team, I am waiting for the appeal result.", "676305": "@lifengnan I believe someone from your team used multiple accounts. " } }