Welcome to the Object Detection Challenge track - a part of the Open Image Challenge! We are excited to present you the large-scale challenge on Open Images V4 dataset recently released by Google AI. In this track, you are asked to tackle the classical computer vision problem of object detection. We are excited to see more sophisticated detection models that will exceed current state-of-the-art performance.
\n\nPlease email us directly on the following topics:
\n\nGood luck!
", "messages": [ { "id": "354271", "postDate": "07/09/2018 08:19:51", "content": "Welcome to the Object Detection Challenge track - a part of the Open Image Challenge! We are excited to present you the large-scale challenge on Open Images V4 dataset recently released by Google AI. In this track, you are asked to tackle the classical computer vision problem of object detection. We are excited to see more sophisticated detection models that will exceed current state-of-the-art performance.
\n\nPlease email us directly on the following topics:
\n\nGood luck!
", "rawMarkdown": "Welcome to the Object Detection Challenge track - a part of the [Open Image Challenge](https://storage.googleapis.com/openimages/web/challenge.html)! We are excited to present you the large-scale challenge on [Open Images V4 dataset](https://storage.googleapis.com/openimages/web/index.html) recently released by [Google AI](https://ai.google/). In this track, you are asked to tackle the classical computer vision problem of object detection. We are excited to see more sophisticated detection models that will exceed current state-of-the-art performance.\n\nPlease email us directly on the following topics:\n\n* Problems with train data and train images download: jponttuset@google.com\n* Question about challenge evaluation metric and its public implementation: oid-challenge2018-contact@google.com\n\nGood luck!", "votes": null }, { "id": "354589", "postDate": "07/09/2018 21:26:10", "content": "As this competition requires huge resources, how about leveling the ground a little by providing some cloud credits?
", "rawMarkdown": "As this competition requires huge resources, how about leveling the ground a little by providing some cloud credits?", "votes": null }, { "id": "354591", "postDate": "07/09/2018 21:29:33", "content": "Hi Moshel, https://www.kaggle.com/c/google-ai-open-images-object-detection-track/discussion/60400#352254
", "rawMarkdown": "Hi Moshel, https://www.kaggle.com/c/google-ai-open-images-object-detection-track/discussion/60400#352254", "votes": null }, { "id": "354598", "postDate": "07/09/2018 22:03:43", "content": "Thank you!
", "rawMarkdown": "Thank you!", "votes": null }, { "id": "355027", "postDate": "07/10/2018 18:08:11", "content": "I am new here, But the way you try and explain your entire effort is really good to see, i hardly see this kind of efforts on the public web space. Thank you.
", "rawMarkdown": "I am new here, But the way you try and explain your entire effort is really good to see, i hardly see this kind of efforts on the public web space. Thank you.", "votes": null }, { "id": "355686", "postDate": "07/12/2018 07:51:26", "content": "Hi，I am a student, I use Mxnet and your public implementation of evaluation used tensorflow.\nSo,how can I evaluate my result in/after training ?\nThank you~
", "rawMarkdown": "Hi，I am a student, I use Mxnet and your public implementation of evaluation used tensorflow.\nSo,how can I evaluate my result in/after training ?\nThank you~", "votes": null }, { "id": "365280", "postDate": "08/02/2018 08:43:55", "content": "Hi all,
\n\nCVDF made zipped files available to download, dividing the train set in 16 files, e.g.:\naws s3 --no-sign-request cp s3://open-images-dataset/tar/train_0.tar.gz [target_dir] (46G)\nLet's hope this makes downloads easier.
\n\nBest,
", "rawMarkdown": "Hi all, \n\nCVDF made zipped files available to download, dividing the train set in 16 files, e.g.:\naws s3 --no-sign-request cp s3://open-images-dataset/tar/train_0.tar.gz [target_dir] (46G)\nLet's hope this makes downloads easier.\n\nBest,", "votes": null }, { "id": "366644", "postDate": "08/06/2018 06:25:51", "content": "Hi, as no one answered my post, i'll ask it again here... @alina is there a way of loading the team coupons into the team leader account? we are two in the team and the training process costs a lot but converging. However, after a few false starts the 500$ are running out. My other team member has a 500$ coupon but I can't use it on my account as its locked into the login credentials. I really don't want to stop the training and build a new instance on his account and try to continue from there.
\n\nThanks,\nMoshe
", "rawMarkdown": "Hi, as no one answered my post, i'll ask it again here... @alina is there a way of loading the team coupons into the team leader account? we are two in the team and the training process costs a lot but converging. However, after a few false starts the 500$ are running out. My other team member has a 500$ coupon but I can't use it on my account as its locked into the login credentials. I really don't want to stop the training and build a new instance on his account and try to continue from there.\n\nThanks,\nMoshe", "votes": null }, { "id": "366835", "postDate": "08/06/2018 16:02:14", "content": "Hi Moshel,
\n\nThe host team unfortunately cannot help you troubleshoot your GCP coupon issues, as Alina and the host team operate separately from that team. I'd recommend contacting Cloud Support to see how you can appropriately transfer coupons..
", "rawMarkdown": "Hi Moshel,\n\nThe host team unfortunately cannot help you troubleshoot your GCP coupon issues, as Alina and the host team operate separately from that team. I'd recommend contacting Cloud Support to see how you can appropriately transfer coupons..", "votes": null }, { "id": "371615", "postDate": "08/17/2018 08:22:15", "content": "I have posted a question on the \"external data thread but got no answer.\nI have read the competition rules several times but couldn't find any reference for disclosure of external data up to the merger date. The only rule is disclosure if you win.\nIs that correct?
", "rawMarkdown": "I have posted a question on the \"external data thread but got no answer.\nI have read the competition rules several times but couldn't find any reference for disclosure of external data up to the merger date. The only rule is disclosure if you win.\nIs that correct?", "votes": null }, { "id": "371779", "postDate": "08/17/2018 15:45:14", "content": "Hi Moshel,
\n\n(I'll copy this to the other thread as well). You must disclose one week in advance, and applies to all participants per the rules, not just if you win. Should you be a winner, and we determine that you had not disclosed the external data one week prior to competition close (merger date), we would disqualify you and move to the next winner.
", "rawMarkdown": "Hi Moshel,\n\n(I'll copy this to the other thread as well). You must disclose one week in advance, and applies to all participants per the rules, not just if you win. Should you be a winner, and we determine that you had not disclosed the external data one week prior to competition close (merger date), we would disqualify you and move to the next winner.", "votes": null }, { "id": "371905", "postDate": "08/17/2018 20:01:49", "content": "Attached is the only paragraph in the rules that addresses external data:
\n\nEXTERNAL DATA\nThe following provision supersedes General Rules Section 7.C. below: “You may use data, other than the Competition Data, as allowed on the Competition Website to develop and test your models and Submissions; provided, you have the right and authority to use such external data for the purposes of the Competition, and to share such data with Sponsor and Kaggle as may be required.\"
\n\nEnglish is not my first language but I can't find in the rules this requirement of disclosure.
", "rawMarkdown": "Attached is the only paragraph in the rules that addresses external data:\n\nEXTERNAL DATA\nThe following provision supersedes General Rules Section 7.C. below: “You may use data, other than the Competition Data, as allowed on the Competition Website to develop and test your models and Submissions; provided, you have the right and authority to use such external data for the purposes of the Competition, and to share such data with Sponsor and Kaggle as may be required.\"\n\nEnglish is not my first language but I can't find in the rules this requirement of disclosure.", "votes": null }, { "id": "371908", "postDate": "08/17/2018 20:02:44", "content": "Hi Moshel,
\n\nDisclosure is included in \"and to share such data with Sponsor and Kaggle as may be required\"
", "rawMarkdown": "Hi Moshel,\n\nDisclosure is included in \"and to share such data with Sponsor and Kaggle as may be required\"", "votes": null }, { "id": "371911", "postDate": "08/17/2018 20:06:38", "content": "This sentence was taken out of context. It starts with: provided, you have the right and authority to use such external data
\n\nSo, if my English is correct, it requires us to have the right to share the data, not to disclose it a week before the competition ends.
\n\nI am sorry for being a pain, but the rules really don't include this requirement, therefore it should be removed
", "rawMarkdown": "This sentence was taken out of context. It starts with: provided, you have the right and authority to use such external data\n\nSo, if my English is correct, it requires us to have the right to share the data, not to disclose it a week before the competition ends.\n\n\nI am sorry for being a pain, but the rules really don't include this requirement, therefore it should be removed", "votes": null }, { "id": "371913", "postDate": "08/17/2018 20:11:29", "content": "Hi Moshel,
\n\nThis wording has been approved by members of the Kaggle Legal team. Should you choose to interpret wording of the rules differently, please note that you may be in violation of the terms and conditions of the competition.
\n\nYou can consider that the external data disclosure thread is covered in the use of \"as may be required\" in the clause you've noted.
", "rawMarkdown": "Hi Moshel,\n\nThis wording has been approved by members of the Kaggle Legal team. Should you choose to interpret wording of the rules differently, please note that you may be in violation of the terms and conditions of the competition. \n\nYou can consider that the external data disclosure thread is covered in the use of \"as may be required\" in the clause you've noted.", "votes": null }, { "id": "371917", "postDate": "08/17/2018 20:17:07", "content": "I have no fight with you, i am just pointing out a mistake.
\n\nThe wording is fine, there is no problem with it. But, it does not require us to disclose external data a week before the finish date. This kind of requirement should appear in the rules. Your interpretation of \"as may he required\" in this context is too broad and won't stand a legal challenge. It is meant for disclosure and sharing after the competition ends.
\n\nI won't comment anymore on this but i suggest you reconsult with the legal team about this specific interpretation of the rules.
", "rawMarkdown": "I have no fight with you, i am just pointing out a mistake.\n\nThe wording is fine, there is no problem with it. But, it does not require us to disclose external data a week before the finish date. This kind of requirement should appear in the rules. Your interpretation of \"as may he required\" in this context is too broad and won't stand a legal challenge. It is meant for disclosure and sharing after the competition ends.\n\nI won't comment anymore on this but i suggest you reconsult with the legal team about this specific interpretation of the rules.", "votes": null }, { "id": "371923", "postDate": "08/17/2018 20:30:35", "content": "Moshel,
\n\nThe intent of the policy is to try and make Kaggle competitions ML competitions, and not simply data collection challenges. We'll reconvene internally on the issue and determine if a language update is appropriate. Thanks for bringing it to our attention!
", "rawMarkdown": "Moshel,\n\nThe intent of the policy is to try and make Kaggle competitions ML competitions, and not simply data collection challenges. We'll reconvene internally on the issue and determine if a language update is appropriate. Thanks for bringing it to our attention!", "votes": null }, { "id": "374891", "postDate": "08/24/2018 05:03:18", "content": "So, the deadline has passed and only about 10 teams submitted. All the others are disqualified?
", "rawMarkdown": "So, the deadline has passed and only about 10 teams submitted. All the others are disqualified?", "votes": null }, { "id": "375098", "postDate": "08/24/2018 15:43:21", "content": "Moshel,
\n\nA team does not have to submit what data sources they used if they were already posted to the forum.
\n\nShould any undisclosed data sources be used, those teams would be disqualified.
", "rawMarkdown": "Moshel,\n\nA team does not have to submit what data sources they used if they were already posted to the forum. \n\nShould any undisclosed data sources be used, those teams would be disqualified.", "votes": null }, { "id": "379192", "postDate": "08/31/2018 01:45:56", "content": "Hi Addison,
\n\nWe highly appreciate the great competition, and it has been a fun and challenging event. I and my teammates have a concern about whether the 1st place team kivajok actually did follow the rules of this competition. We appreciate your sincere comments on the following points.
\n\nIn the competition, it was clearly stated that [1]
\n\n\n\n\nPlease use this thread to post any external data and pre-trained models you use for your\n solution. Reminder, disclosure is required one week prior to the submission deadline.
\n
In our team, we followed this rule [2], whereas team kivajok did not disclose their external data and pretrained models until Aug 29. [3]
\n\nWhile here you state that
\n\n\n\n\nA team does not have to submit what data sources they used if they were already posted to the forum.\n Should any undisclosed data sources be used, those teams would be disqualified.
\n
we consider that team kivajok is still breaking the rules, since they claim that they used pretrained models such as deformable convnets [4], which was not disclosed by any other team within the external data disclosure deadline, and is therefore subject to disqualification in this competition. The final scores were remarkably close among the top teams so even if their advantage was tiny, the result might have been different had they followed the rules properly.
\n\nWe regard it lack of sportsmanship and fairness that a team that did not follow the rules gets an advantage by not following the external data disclosure deadline, and thereby advancing on the leaderboard using more information than other teams. Our team and many other teams played by fair rules in this competition and we sincerely hope that the organizers will adhere to the strict rules that were set in this competition.
\n\nHi Iwiw,
\n\nThanks for bringing it to our attention. Should you ever feel that another competitor violated the rules, I'd encourage you to submit a report to compliance@kaggle.com.
\n\nThe rules are between the Sponsor (Google AI) and the Kaggle community. I'll bring this to their attention, and they will be able to review the winning solution to determine if it violates the rules of the competition. Those who place \"in the money\" on the leaderboard are considered potential winners until the hosts have the opportunity to review the winning models.
\n\nThanks,
\n\nAddison
", "rawMarkdown": "Hi Iwiw,\n\nThanks for bringing it to our attention. Should you ever feel that another competitor violated the rules, I'd encourage you to submit a report to compliance@kaggle.com.\n\nThe rules are between the Sponsor (Google AI) and the Kaggle community. I'll bring this to their attention, and they will be able to review the winning solution to determine if it violates the rules of the competition. Those who place \"in the money\" on the leaderboard are considered *potential* winners until the hosts have the opportunity to review the winning models.\n\nThanks,\n\nAddison", "votes": null } ], "comments": [ { "id": 354589, "author_name": "moshel", "author_url": "", "post_date": "07/09/2018 21:26:10", "content": "As this competition requires huge resources, how about leveling the ground a little by providing some cloud credits?
", "votes": null, "replies": [ { "id": 354591, "author_name": "addisonhoward", "author_url": "", "post_date": "07/09/2018 21:29:33", "content": "Hi Moshel, https://www.kaggle.com/c/google-ai-open-images-object-detection-track/discussion/60400#352254
", "votes": null, "replies": [] }, { "id": 354598, "author_name": "moshel", "author_url": "", "post_date": "07/09/2018 22:03:43", "content": "Thank you!
", "votes": null, "replies": [] }, { "id": 366644, "author_name": "moshel", "author_url": "", "post_date": "08/06/2018 06:25:51", "content": "Hi, as no one answered my post, i'll ask it again here... @alina is there a way of loading the team coupons into the team leader account? we are two in the team and the training process costs a lot but converging. However, after a few false starts the 500$ are running out. My other team member has a 500$ coupon but I can't use it on my account as its locked into the login credentials. I really don't want to stop the training and build a new instance on his account and try to continue from there.
\n\nThanks,\nMoshe
", "votes": null, "replies": [] }, { "id": 366835, "author_name": "addisonhoward", "author_url": "", "post_date": "08/06/2018 16:02:14", "content": "Hi Moshel,
\n\nThe host team unfortunately cannot help you troubleshoot your GCP coupon issues, as Alina and the host team operate separately from that team. I'd recommend contacting Cloud Support to see how you can appropriately transfer coupons..
", "votes": null, "replies": [] } ] }, { "id": 355027, "author_name": "vishalprajapati", "author_url": "", "post_date": "07/10/2018 18:08:11", "content": "I am new here, But the way you try and explain your entire effort is really good to see, i hardly see this kind of efforts on the public web space. Thank you.
", "votes": null, "replies": [] }, { "id": 355686, "author_name": "mayufeng", "author_url": "", "post_date": "07/12/2018 07:51:26", "content": "Hi，I am a student, I use Mxnet and your public implementation of evaluation used tensorflow.\nSo,how can I evaluate my result in/after training ?\nThank you~
", "votes": null, "replies": [] }, { "id": 365280, "author_name": "jponttusset", "author_url": "", "post_date": "08/02/2018 08:43:55", "content": "Hi all,
\n\nCVDF made zipped files available to download, dividing the train set in 16 files, e.g.:\naws s3 --no-sign-request cp s3://open-images-dataset/tar/train_0.tar.gz [target_dir] (46G)\nLet's hope this makes downloads easier.
\n\nBest,
", "votes": null, "replies": [] }, { "id": 371615, "author_name": "moshel", "author_url": "", "post_date": "08/17/2018 08:22:15", "content": "I have posted a question on the \"external data thread but got no answer.\nI have read the competition rules several times but couldn't find any reference for disclosure of external data up to the merger date. The only rule is disclosure if you win.\nIs that correct?
", "votes": null, "replies": [ { "id": 371779, "author_name": "addisonhoward", "author_url": "", "post_date": "08/17/2018 15:45:14", "content": "Hi Moshel,
\n\n(I'll copy this to the other thread as well). You must disclose one week in advance, and applies to all participants per the rules, not just if you win. Should you be a winner, and we determine that you had not disclosed the external data one week prior to competition close (merger date), we would disqualify you and move to the next winner.
", "votes": null, "replies": [] }, { "id": 371905, "author_name": "moshel", "author_url": "", "post_date": "08/17/2018 20:01:49", "content": "Attached is the only paragraph in the rules that addresses external data:
\n\nEXTERNAL DATA\nThe following provision supersedes General Rules Section 7.C. below: “You may use data, other than the Competition Data, as allowed on the Competition Website to develop and test your models and Submissions; provided, you have the right and authority to use such external data for the purposes of the Competition, and to share such data with Sponsor and Kaggle as may be required.\"
\n\nEnglish is not my first language but I can't find in the rules this requirement of disclosure.
", "votes": null, "replies": [] }, { "id": 371908, "author_name": "addisonhoward", "author_url": "", "post_date": "08/17/2018 20:02:44", "content": "Hi Moshel,
\n\nDisclosure is included in \"and to share such data with Sponsor and Kaggle as may be required\"
", "votes": null, "replies": [] }, { "id": 371911, "author_name": "moshel", "author_url": "", "post_date": "08/17/2018 20:06:38", "content": "This sentence was taken out of context. It starts with: provided, you have the right and authority to use such external data
\n\nSo, if my English is correct, it requires us to have the right to share the data, not to disclose it a week before the competition ends.
\n\nI am sorry for being a pain, but the rules really don't include this requirement, therefore it should be removed
", "votes": null, "replies": [] }, { "id": 371913, "author_name": "addisonhoward", "author_url": "", "post_date": "08/17/2018 20:11:29", "content": "Hi Moshel,
\n\nThis wording has been approved by members of the Kaggle Legal team. Should you choose to interpret wording of the rules differently, please note that you may be in violation of the terms and conditions of the competition.
\n\nYou can consider that the external data disclosure thread is covered in the use of \"as may be required\" in the clause you've noted.
", "votes": null, "replies": [] }, { "id": 371917, "author_name": "moshel", "author_url": "", "post_date": "08/17/2018 20:17:07", "content": "I have no fight with you, i am just pointing out a mistake.
\n\nThe wording is fine, there is no problem with it. But, it does not require us to disclose external data a week before the finish date. This kind of requirement should appear in the rules. Your interpretation of \"as may he required\" in this context is too broad and won't stand a legal challenge. It is meant for disclosure and sharing after the competition ends.
\n\nI won't comment anymore on this but i suggest you reconsult with the legal team about this specific interpretation of the rules.
", "votes": null, "replies": [] }, { "id": 371923, "author_name": "addisonhoward", "author_url": "", "post_date": "08/17/2018 20:30:35", "content": "Moshel,
\n\nThe intent of the policy is to try and make Kaggle competitions ML competitions, and not simply data collection challenges. We'll reconvene internally on the issue and determine if a language update is appropriate. Thanks for bringing it to our attention!
", "votes": null, "replies": [] }, { "id": 374891, "author_name": "moshel", "author_url": "", "post_date": "08/24/2018 05:03:18", "content": "So, the deadline has passed and only about 10 teams submitted. All the others are disqualified?
", "votes": null, "replies": [] }, { "id": 375098, "author_name": "addisonhoward", "author_url": "", "post_date": "08/24/2018 15:43:21", "content": "Moshel,
\n\nA team does not have to submit what data sources they used if they were already posted to the forum.
\n\nShould any undisclosed data sources be used, those teams would be disqualified.
", "votes": null, "replies": [] }, { "id": 379192, "author_name": "takiba", "author_url": "", "post_date": "08/31/2018 01:45:56", "content": "Hi Addison,
\n\nWe highly appreciate the great competition, and it has been a fun and challenging event. I and my teammates have a concern about whether the 1st place team kivajok actually did follow the rules of this competition. We appreciate your sincere comments on the following points.
\n\nIn the competition, it was clearly stated that [1]
\n\n\n\n\nPlease use this thread to post any external data and pre-trained models you use for your\n solution. Reminder, disclosure is required one week prior to the submission deadline.
\n
In our team, we followed this rule [2], whereas team kivajok did not disclose their external data and pretrained models until Aug 29. [3]
\n\nWhile here you state that
\n\n\n\n\nA team does not have to submit what data sources they used if they were already posted to the forum.\n Should any undisclosed data sources be used, those teams would be disqualified.
\n
we consider that team kivajok is still breaking the rules, since they claim that they used pretrained models such as deformable convnets [4], which was not disclosed by any other team within the external data disclosure deadline, and is therefore subject to disqualification in this competition. The final scores were remarkably close among the top teams so even if their advantage was tiny, the result might have been different had they followed the rules properly.
\n\nWe regard it lack of sportsmanship and fairness that a team that did not follow the rules gets an advantage by not following the external data disclosure deadline, and thereby advancing on the leaderboard using more information than other teams. Our team and many other teams played by fair rules in this competition and we sincerely hope that the organizers will adhere to the strict rules that were set in this competition.
\n\nHi Iwiw,
\n\nThanks for bringing it to our attention. Should you ever feel that another competitor violated the rules, I'd encourage you to submit a report to compliance@kaggle.com.
\n\nThe rules are between the Sponsor (Google AI) and the Kaggle community. I'll bring this to their attention, and they will be able to review the winning solution to determine if it violates the rules of the competition. Those who place \"in the money\" on the leaderboard are considered potential winners until the hosts have the opportunity to review the winning models.
\n\nThanks,
\n\nAddison
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "354271": "Welcome to the Object Detection Challenge track - a part of the [Open Image Challenge](https://storage.googleapis.com/openimages/web/challenge.html)! We are excited to present you the large-scale challenge on [Open Images V4 dataset](https://storage.googleapis.com/openimages/web/index.html) recently released by [Google AI](https://ai.google/). In this track, you are asked to tackle the classical computer vision problem of object detection. We are excited to see more sophisticated detection models that will exceed current state-of-the-art performance.\n\nPlease email us directly on the following topics:\n\n* Problems with train data and train images download: jponttuset@google.com\n* Question about challenge evaluation metric and its public implementation: oid-challenge2018-contact@google.com\n\nGood luck!", "354589": "As this competition requires huge resources, how about leveling the ground a little by providing some cloud credits?", "354591": "Hi Moshel, https://www.kaggle.com/c/google-ai-open-images-object-detection-track/discussion/60400#352254", "354598": "Thank you!", "355027": "I am new here, But the way you try and explain your entire effort is really good to see, i hardly see this kind of efforts on the public web space. Thank you.", "355686": "Hi，I am a student, I use Mxnet and your public implementation of evaluation used tensorflow.\nSo,how can I evaluate my result in/after training ?\nThank you~", "365280": "Hi all, \n\nCVDF made zipped files available to download, dividing the train set in 16 files, e.g.:\naws s3 --no-sign-request cp s3://open-images-dataset/tar/train_0.tar.gz [target_dir] (46G)\nLet's hope this makes downloads easier.\n\nBest,", "366644": "Hi, as no one answered my post, i'll ask it again here... @alina is there a way of loading the team coupons into the team leader account? we are two in the team and the training process costs a lot but converging. However, after a few false starts the 500$ are running out. My other team member has a 500$ coupon but I can't use it on my account as its locked into the login credentials. I really don't want to stop the training and build a new instance on his account and try to continue from there.\n\nThanks,\nMoshe", "366835": "Hi Moshel,\n\nThe host team unfortunately cannot help you troubleshoot your GCP coupon issues, as Alina and the host team operate separately from that team. I'd recommend contacting Cloud Support to see how you can appropriately transfer coupons..", "371615": "I have posted a question on the \"external data thread but got no answer.\nI have read the competition rules several times but couldn't find any reference for disclosure of external data up to the merger date. The only rule is disclosure if you win.\nIs that correct?", "371779": "Hi Moshel,\n\n(I'll copy this to the other thread as well). You must disclose one week in advance, and applies to all participants per the rules, not just if you win. Should you be a winner, and we determine that you had not disclosed the external data one week prior to competition close (merger date), we would disqualify you and move to the next winner.", "371905": "Attached is the only paragraph in the rules that addresses external data:\n\nEXTERNAL DATA\nThe following provision supersedes General Rules Section 7.C. below: “You may use data, other than the Competition Data, as allowed on the Competition Website to develop and test your models and Submissions; provided, you have the right and authority to use such external data for the purposes of the Competition, and to share such data with Sponsor and Kaggle as may be required.\"\n\nEnglish is not my first language but I can't find in the rules this requirement of disclosure.", "371908": "Hi Moshel,\n\nDisclosure is included in \"and to share such data with Sponsor and Kaggle as may be required\"", "371911": "This sentence was taken out of context. It starts with: provided, you have the right and authority to use such external data\n\nSo, if my English is correct, it requires us to have the right to share the data, not to disclose it a week before the competition ends.\n\n\nI am sorry for being a pain, but the rules really don't include this requirement, therefore it should be removed", "371913": "Hi Moshel,\n\nThis wording has been approved by members of the Kaggle Legal team. Should you choose to interpret wording of the rules differently, please note that you may be in violation of the terms and conditions of the competition. \n\nYou can consider that the external data disclosure thread is covered in the use of \"as may be required\" in the clause you've noted.", "371917": "I have no fight with you, i am just pointing out a mistake.\n\nThe wording is fine, there is no problem with it. But, it does not require us to disclose external data a week before the finish date. This kind of requirement should appear in the rules. Your interpretation of \"as may he required\" in this context is too broad and won't stand a legal challenge. It is meant for disclosure and sharing after the competition ends.\n\nI won't comment anymore on this but i suggest you reconsult with the legal team about this specific interpretation of the rules.", "371923": "Moshel,\n\nThe intent of the policy is to try and make Kaggle competitions ML competitions, and not simply data collection challenges. We'll reconvene internally on the issue and determine if a language update is appropriate. Thanks for bringing it to our attention!", "374891": "So, the deadline has passed and only about 10 teams submitted. All the others are disqualified?", "375098": "Moshel,\n\nA team does not have to submit what data sources they used if they were already posted to the forum. \n\nShould any undisclosed data sources be used, those teams would be disqualified.", "379192": "Hi Addison,\n\nWe highly appreciate the great competition, and it has been a fun and challenging event. I and my teammates have a concern about whether the 1st place team kivajok actually did follow the rules of this competition. We appreciate your sincere comments on the following points.\n\nIn the competition, it was clearly stated that [1]\n\n> Please use this thread to post any external data and pre-trained models you use for your\n> solution. Reminder, disclosure is required one week prior to the submission deadline.\n\nIn our team, we followed this rule [2], whereas team kivajok did not disclose their external data and pretrained models until Aug 29. [3]\n\nWhile here you state that\n\n> A team does not have to submit what data sources they used if they were already posted to the forum.\n> Should any undisclosed data sources be used, those teams would be disqualified.\n\nwe consider that team kivajok is still breaking the rules, since they claim that they used pretrained models such as deformable convnets [4], which was not disclosed by any other team within the external data disclosure deadline, and is therefore subject to disqualification in this competition. The final scores were remarkably close among the top teams so even if their advantage was tiny, the result might have been different had they followed the rules properly.\n\nWe regard it lack of sportsmanship and fairness that a team that did not follow the rules gets an advantage by not following the external data disclosure deadline, and thereby advancing on the leaderboard using more information than other teams. Our team and many other teams played by fair rules in this competition and we sincerely hope that the organizers will adhere to the strict rules that were set in this competition.\n\n - [1] https://www.kaggle.com/c/google-ai-open-images-object-detection-track/discussion/62741\n - [2] https://www.kaggle.com/c/google-ai-open-images-object-detection-track/discussion/62741#374556\n - [3] https://www.kaggle.com/c/google-ai-open-images-object-detection-track/discussion/62741#377600\n - [4] https://github.com/msracver/Deformable-ConvNets", "379194": "Hi Iwiw,\n\nThanks for bringing it to our attention. Should you ever feel that another competitor violated the rules, I'd encourage you to submit a report to compliance@kaggle.com.\n\nThe rules are between the Sponsor (Google AI) and the Kaggle community. I'll bring this to their attention, and they will be able to review the winning solution to determine if it violates the rules of the competition. Those who place \"in the money\" on the leaderboard are considered *potential* winners until the hosts have the opportunity to review the winning models.\n\nThanks,\n\nAddison" }, "source": "meta" }