Hi all!
\n\nGetting a bit emotional, and therefore I had to write.
\n\nThis competition was a very special and emotional one for me on many levels. I am sure many other Kagglers, especially, who dared to take part in this competition in spite of having limited computing resources can relate as well! This is my first serious Kaggle competition, not an easy one to start for sure, but the amount of learning was so huge that I will never regret it.
\n\nThanks to @unkownhihi and his team, for establishing themselves as public figures from early on who have selflessly helped so many Kagglers like me to start the competition with a simple public kernel, throwing away so many hints in discussions, and for continuing to share extremely helpful kernels. You are the best!
\n\nThanks to @humananalog, for that controversial kernel, that in fact turned out to be one of the most helpful notebooks that gave away hints to many of us on how to successfully build a working dataset, train a transfer learning model to cross a bubble (that at some point of time I thought I will not be able to cross) around the log loss of 0.69LB. Kudos!
\n\nThanks to everyone who contributed to the discussions- a place that literally becomes like a social media timeline, that I used to check when I am having my breakfast, dinner, lunch or when I am stuck in the subway. Literally, we all read it for entertainment.
\n\nThanks to GCP, Amazon for timely cloud resources. Thanks to Kaggle for your super fast GPU (and UI modifications)! Without you, many of us would have been drifting without any directions.
\n\nAnother thing that will make me never forget this competition is the unfortunate Covid-19 situation. We all suffered this way or that, but the competition kept going, showing science wins no matter what!
\n\nIt all started in 2019, a year when social distancing was a concept that I never heard, and now writing it from my room, while I am self-isolated. No matter what happens next, as someone said, \"It's All About the Journey, Not the Destination\".
\n\nIndeed, what a journey!
\n\nThanks, everyone! All the best!
", "messages": [ { "id": "792416", "postDate": "03/31/2020 07:21:35", "content": "Hi all!
\n\nGetting a bit emotional, and therefore I had to write.
\n\nThis competition was a very special and emotional one for me on many levels. I am sure many other Kagglers, especially, who dared to take part in this competition in spite of having limited computing resources can relate as well! This is my first serious Kaggle competition, not an easy one to start for sure, but the amount of learning was so huge that I will never regret it.
\n\nThanks to @unkownhihi and his team, for establishing themselves as public figures from early on who have selflessly helped so many Kagglers like me to start the competition with a simple public kernel, throwing away so many hints in discussions, and for continuing to share extremely helpful kernels. You are the best!
\n\nThanks to @humananalog, for that controversial kernel, that in fact turned out to be one of the most helpful notebooks that gave away hints to many of us on how to successfully build a working dataset, train a transfer learning model to cross a bubble (that at some point of time I thought I will not be able to cross) around the log loss of 0.69LB. Kudos!
\n\nThanks to everyone who contributed to the discussions- a place that literally becomes like a social media timeline, that I used to check when I am having my breakfast, dinner, lunch or when I am stuck in the subway. Literally, we all read it for entertainment.
\n\nThanks to GCP, Amazon for timely cloud resources. Thanks to Kaggle for your super fast GPU (and UI modifications)! Without you, many of us would have been drifting without any directions.
\n\nAnother thing that will make me never forget this competition is the unfortunate Covid-19 situation. We all suffered this way or that, but the competition kept going, showing science wins no matter what!
\n\nIt all started in 2019, a year when social distancing was a concept that I never heard, and now writing it from my room, while I am self-isolated. No matter what happens next, as someone said, \"It's All About the Journey, Not the Destination\".
\n\nIndeed, what a journey!
\n\nThanks, everyone! All the best!
", "rawMarkdown": "Hi all!\n\nGetting a bit emotional, and therefore I had to write.\n\nThis competition was a very special and emotional one for me on many levels. I am sure many other Kagglers, especially, who dared to take part in this competition in spite of having limited computing resources can relate as well! This is my first serious Kaggle competition, not an easy one to start for sure, but the amount of learning was so huge that I will never regret it.\n\nThanks to @unkownhihi and his team, for establishing themselves as public figures from early on who have selflessly helped so many Kagglers like me to start the competition with a simple public kernel, throwing away so many hints in discussions, and for continuing to share extremely helpful kernels. You are the best!\n\nThanks to @humananalog, for that controversial kernel, that in fact turned out to be one of the most helpful notebooks that gave away hints to many of us on how to successfully build a working dataset, train a transfer learning model to cross a bubble (that at some point of time I thought I will not be able to cross) around the log loss of 0.69LB. Kudos!\n\nThanks to everyone who contributed to the discussions- a place that literally becomes like a social media timeline, that I used to check when I am having my breakfast, dinner, lunch or when I am stuck in the subway. Literally, we all read it for entertainment.\n\nThanks to GCP, Amazon for timely cloud resources. Thanks to Kaggle for your super fast GPU (and UI modifications)! Without you, many of us would have been drifting without any directions.\n\nAnother thing that will make me never forget this competition is the unfortunate Covid-19 situation. We all suffered this way or that, but the competition kept going, showing science wins no matter what!\n\nIt all started in 2019, a year when social distancing was a concept that I never heard, and now writing it from my room, while I am self-isolated. No matter what happens next, as someone said, \"It's All About the Journey, Not the Destination\".\n\nIndeed, what a journey!\n\nThanks, everyone! All the best!", "votes": null }, { "id": "792439", "postDate": "03/31/2020 07:41:26", "content": "May the force be with you ! I will keep my emotion for the potential big drop tomorrow π
", "rawMarkdown": "May the force be with you ! I will keep my emotion for the potential big drop tomorrow π", "votes": null }, { "id": "792442", "postDate": "03/31/2020 07:46:15", "content": "My final submission is running, so I donβt want any emotion or stress tomorrow :D Anyway, I believe it will take some time to run these submission in private set as this competition may be a special case. Also, they will give public score only for 2 submissions, thank god thatβs a relief from the panic for not submitting other potential good submissions that we donβt select.
", "rawMarkdown": "My final submission is running, so I donβt want any emotion or stress tomorrow :D Anyway, I believe it will take some time to run these submission in private set as this competition may be a special case. Also, they will give public score only for 2 submissions, thank god thatβs a relief from the panic for not submitting other potential good submissions that we donβt select.", "votes": null }, { "id": "792469", "postDate": "03/31/2020 08:18:51", "content": "haha, you are right. It will take time to score, so at least I will cry another day :)
", "rawMarkdown": "haha, you are right. It will take time to score, so at least I will cry another day :)", "votes": null }, { "id": "792563", "postDate": "03/31/2020 10:30:08", "content": "Well put, I relate on many levels to what you said esp it being my first. Definitely a bittersweet end to a great competition.
", "rawMarkdown": "Well put, I relate on many levels to what you said esp it being my first. Definitely a bittersweet end to a great competition.", "votes": null }, { "id": "792902", "postDate": "03/31/2020 16:23:56", "content": "Yes, indeed, a great journey. Many ups and downs, surprises(good and bad), discoveries, new thoughts. Good luck at the private leaderboard to all!!!
", "rawMarkdown": "Yes, indeed, a great journey. Many ups and downs, surprises(good and bad), discoveries, new thoughts. Good luck at the private leaderboard to all!!!", "votes": null }, { "id": "792904", "postDate": "03/31/2020 16:26:09", "content": "@unknownhihi \nwhat do you think about the shakeup?
", "rawMarkdown": "unknownhihi \nwhat do you think about the shakeup?", "votes": null }, { "id": "792920", "postDate": "03/31/2020 16:36:55", "content": "Well, ensemble as many model as you can, do Kfold if possible. Stress test your model(but probably a little bit too late)
\n\nWe, unfortunately, ran out of aws quota and had to terminate all the instances(which means permanently delete) that have our model on it. π£ Now the only thing we could do is cross our fingers.
", "rawMarkdown": "Well, ensemble as many model as you can, do Kfold if possible. Stress test your model(but probably a little bit too late)\n\nWe, unfortunately, ran out of aws quota and had to terminate all the instances(which means permanently delete) that have our model on it. π£ Now the only thing we could do is cross our fingers.", "votes": null }, { "id": "792943", "postDate": "03/31/2020 17:09:30", "content": "\n\n\n\n\nWell, ensemble as many model as you can, do Kfold if possible. Stress test your model(but probably a little bit too late)
\n
Done that, fingers crossed!
", "rawMarkdown": ">> Well, ensemble as many model as you can, do Kfold if possible. Stress test your model(but probably a little bit too late)\n\nDone that, fingers crossed!", "votes": null }, { "id": "792948", "postDate": "03/31/2020 17:17:23", "content": "Not sure I understood correctly your words about the big drop. There will be no drop tomorrow because the results will be revealed in the end of April.
\n\n\n", "rawMarkdown": "Not sure I understood correctly your words about the big drop. There will be no drop tomorrow because the results will be revealed in the end of April.\n\n> (Estimated) April 22, 2020 - Private leaderboard revealed.\n\n>Note: Because the private test set is held outside of Kaggle's platform, the private leaderboard will >not be revealed immediately. We have estimated the scoring timeframe, but this is subject to revision.", "votes": null }, { "id": "792952", "postDate": "03/31/2020 17:20:33", "content": "(Estimated) April 22, 2020 - Private leaderboard revealed.
\n \nNote: Because the private test set is held outside of Kaggle's platform, the private leaderboard will >not be revealed immediately. We have estimated the scoring timeframe, but this is subject to revision.
\n
Yes, great journey! 2 last submissions in progress... The competition is over and I've learnt a lot once again. Kaggle is really a great place. Thanks to my motivated teammate @tikutiku that did great job with many models experiments and validation scheme totally different than mine that made our ensemble efficient (at least on public LB). Also, thanks to @timesler for his great facenet-pytorch MTCNN detector which we've used for most of our models. @timesler : Add TPU support and your lib will be the best π . Thanks to @humananalog that gives a boost to the competition by pushing us to do better with his kernel. Thanks to the organizers @juliaelliott for the TPU and GCP coupons. I've trained my first TPU (pytorch-XLA) model with this competition, it's really impressive with multi-cores, it's so fast.
\n\nThanks to all competitors for sharing idea and kernels. I can't wait to see solutions and approaches from top competitors. How can they reach 0.2? How did they deal with high quality fakes? How far their solution from State Of The Art? And finally how this competition will really help to solve the problem. If you look at https://www.thispersondoesnotexist.com/ it's a bit scary how great it is.
\n\nResults within the next weeks, shake up expected. βοΈ
", "rawMarkdown": "Yes, great journey! 2 last submissions in progress... The competition is over and I've learnt a lot once again. Kaggle is really a great place. Thanks to my motivated teammate @tikutiku that did great job with many models experiments and validation scheme totally different than mine that made our ensemble efficient (at least on public LB). Also, thanks to @timesler for his great facenet-pytorch MTCNN detector which we've used for most of our models. @timesler : Add TPU support and your lib will be the best π . Thanks to @humananalog that gives a boost to the competition by pushing us to do better with his kernel. Thanks to the organizers @juliaelliott for the TPU and GCP coupons. I've trained my first TPU (pytorch-XLA) model with this competition, it's really impressive with multi-cores, it's so fast.\n\nThanks to all competitors for sharing idea and kernels. I can't wait to see solutions and approaches from top competitors. How can they reach 0.2? How did they deal with high quality fakes? How far their solution from State Of The Art? And finally how this competition will really help to solve the problem. If you look at https://www.thispersondoesnotexist.com/ it's a bit scary how great it is.\n\nResults within the next weeks, shake up expected. βοΈ", "votes": null }, { "id": "792975", "postDate": "03/31/2020 17:35:07", "content": "That's a long time. Time to focus on M5 :D
", "rawMarkdown": "That's a long time. Time to focus on M5 :D", "votes": null }, { "id": "793099", "postDate": "03/31/2020 19:51:01", "content": "That's what I was going to say as well. :) \nTime to focus on other competitions while top solutions are being scored. ;)\nBest of luck for all of you!
", "rawMarkdown": "That's what I was going to say as well. :) \nTime to focus on other competitions while top solutions are being scored. ;)\nBest of luck for all of you!", "votes": null }, { "id": "793103", "postDate": "03/31/2020 19:57:31", "content": "Yes, what a journey. This is also my first CV competition. Many thanks for the resources provided by AWS and GCP. Without their support, it's really hard to keep on this competition for a student who does not have great equipment.\nAnd I really appreciate the field of computer vision after the long learning journey. I have learnt a lot from the wonderful public notebook and discussions. Thanks for all competitors who share their ideas, especially @unkownhihi π .\nI can't wait to see the best solution and see whether it can really protect us from deepfake attack in real world!
", "rawMarkdown": "Yes, what a journey. This is also my first CV competition. Many thanks for the resources provided by AWS and GCP. Without their support, it's really hard to keep on this competition for a student who does not have great equipment.\nAnd I really appreciate the field of computer vision after the long learning journey. I have learnt a lot from the wonderful public notebook and discussions. Thanks for all competitors who share their ideas, especially @unkownhihi π .\nI can't wait to see the best solution and see whether it can really protect us from deepfake attack in real world!", "votes": null }, { "id": "793118", "postDate": "03/31/2020 20:15:11", "content": "I feel the same. It was a great journey. Everyone was helpful. Special thanks to @unkownhihi.
", "rawMarkdown": "I feel the same. It was a great journey. Everyone was helpful. Special thanks to @unkownhihi.", "votes": null }, { "id": "793152", "postDate": "03/31/2020 20:49:39", "content": "I really enjoyed the tournament. Yet, the competition was to real and I was only for learning :)
", "rawMarkdown": "I really enjoyed the tournament. Yet, the competition was to real and I was only for learning :)", "votes": null }, { "id": "793177", "postDate": "03/31/2020 21:17:54", "content": "I'm quite new here, but it's very exciting to join this platform. Thanks for sharing!
", "rawMarkdown": "I'm quite new here, but it's very exciting to join this platform. Thanks for sharing!", "votes": null }, { "id": "793195", "postDate": "03/31/2020 21:49:05", "content": "2 final submissions selected. Good luck to everyone!
", "rawMarkdown": "2 final submissions selected. Good luck to everyone!", "votes": null }, { "id": "793241", "postDate": "03/31/2020 22:38:41", "content": "Indeed, amazing competition. Now I know that a dataset of ~450 GB of videos is not that scary as it seems :)
", "rawMarkdown": "Indeed, amazing competition. Now I know that a dataset of ~450 GB of videos is not that scary as it seems :)", "votes": null }, { "id": "793305", "postDate": "03/31/2020 23:55:07", "content": "LOL Added more dramatic! We got our new score 10 minutes before deadline..... π¬
", "rawMarkdown": "LOL Added more dramatic! We got our new score 10 minutes before deadline..... π¬", "votes": null }, { "id": "793473", "postDate": "04/01/2020 03:54:43", "content": "Looking forward to my electricity bill (not!) π₯π₯
", "rawMarkdown": "Looking forward to my electricity bill (not!) π₯π₯", "votes": null }, { "id": "794370", "postDate": "04/01/2020 19:04:58", "content": "Thank you for your strength and courage! our words help all of us in isolation to get through tough times.\nKeep up the good work!
", "rawMarkdown": "Thank you for your strength and courage! our words help all of us in isolation to get through tough times.\nKeep up the good work!", "votes": null }, { "id": "794853", "postDate": "04/02/2020 06:25:58", "content": "Just getting started with this exciting journey towards DS, AI, DL ....
", "rawMarkdown": "Just getting started with this exciting journey towards DS, AI, DL ....", "votes": null }, { "id": "795562", "postDate": "04/02/2020 20:34:22", "content": "Encouraging words!
", "rawMarkdown": "Encouraging words!", "votes": null }, { "id": "796605", "postDate": "04/03/2020 18:28:07", "content": "Dear Fellow Kagglers,\nThis is my first comment here as I am starting my journey on Kaggle. I would appreciate all you guidance to make it through and work along withyou all.\nThanks
", "rawMarkdown": "Dear Fellow Kagglers,\nThis is my first comment here as I am starting my journey on Kaggle. I would appreciate all you guidance to make it through and work along withyou all.\nThanks", "votes": null }, { "id": "804856", "postDate": "04/12/2020 03:42:02", "content": "Thank you for inspiring me, awesome story
", "rawMarkdown": "Thank you for inspiring me, awesome story", "votes": null }, { "id": "806600", "postDate": "04/13/2020 21:44:20", "content": "Fantastic achievement. Truly inspiring.
", "rawMarkdown": "Fantastic achievement. Truly inspiring.", "votes": null }, { "id": "807519", "postDate": "04/14/2020 17:38:44", "content": "congrats
", "rawMarkdown": "congrats", "votes": null }, { "id": "812634", "postDate": "04/18/2020 21:37:03", "content": "encouraging words... π
", "rawMarkdown": "encouraging words... π", "votes": null }, { "id": "818133", "postDate": "04/23/2020 16:35:55", "content": "π
", "rawMarkdown": "π", "votes": null } ], "comments": [ { "id": 792439, "author_name": "nyleve", "author_url": "", "post_date": "03/31/2020 07:41:26", "content": "May the force be with you ! I will keep my emotion for the potential big drop tomorrow π
", "votes": null, "replies": [ { "id": 792442, "author_name": "debanga", "author_url": "", "post_date": "03/31/2020 07:46:15", "content": "My final submission is running, so I donβt want any emotion or stress tomorrow :D Anyway, I believe it will take some time to run these submission in private set as this competition may be a special case. Also, they will give public score only for 2 submissions, thank god thatβs a relief from the panic for not submitting other potential good submissions that we donβt select.
", "votes": null, "replies": [] }, { "id": 792469, "author_name": "nyleve", "author_url": "", "post_date": "03/31/2020 08:18:51", "content": "haha, you are right. It will take time to score, so at least I will cry another day :)
", "votes": null, "replies": [] }, { "id": 792948, "author_name": "vpaslay", "author_url": "", "post_date": "03/31/2020 17:17:23", "content": "Not sure I understood correctly your words about the big drop. There will be no drop tomorrow because the results will be revealed in the end of April.
\n\n\n", "votes": null, "replies": [] }, { "id": 792975, "author_name": "debanga", "author_url": "", "post_date": "03/31/2020 17:35:07", "content": "(Estimated) April 22, 2020 - Private leaderboard revealed.
\n \nNote: Because the private test set is held outside of Kaggle's platform, the private leaderboard will >not be revealed immediately. We have estimated the scoring timeframe, but this is subject to revision.
\n
That's a long time. Time to focus on M5 :D
", "votes": null, "replies": [] }, { "id": 793099, "author_name": "yassinealouini", "author_url": "", "post_date": "03/31/2020 19:51:01", "content": "That's what I was going to say as well. :) \nTime to focus on other competitions while top solutions are being scored. ;)\nBest of luck for all of you!
", "votes": null, "replies": [] } ] }, { "id": 792563, "author_name": "shinig4mi", "author_url": "", "post_date": "03/31/2020 10:30:08", "content": "Well put, I relate on many levels to what you said esp it being my first. Definitely a bittersweet end to a great competition.
", "votes": null, "replies": [] }, { "id": 792902, "author_name": "unkownhihi", "author_url": "", "post_date": "03/31/2020 16:23:56", "content": "Yes, indeed, a great journey. Many ups and downs, surprises(good and bad), discoveries, new thoughts. Good luck at the private leaderboard to all!!!
", "votes": null, "replies": [ { "id": 792904, "author_name": "nxrprime", "author_url": "", "post_date": "03/31/2020 16:26:09", "content": "@unknownhihi \nwhat do you think about the shakeup?
", "votes": null, "replies": [] }, { "id": 792920, "author_name": "unkownhihi", "author_url": "", "post_date": "03/31/2020 16:36:55", "content": "Well, ensemble as many model as you can, do Kfold if possible. Stress test your model(but probably a little bit too late)
\n\nWe, unfortunately, ran out of aws quota and had to terminate all the instances(which means permanently delete) that have our model on it. π£ Now the only thing we could do is cross our fingers.
", "votes": null, "replies": [] }, { "id": 792943, "author_name": "debanga", "author_url": "", "post_date": "03/31/2020 17:09:30", "content": "\n\n\n\n\nWell, ensemble as many model as you can, do Kfold if possible. Stress test your model(but probably a little bit too late)
\n
Done that, fingers crossed!
", "votes": null, "replies": [] }, { "id": 793195, "author_name": "mpware", "author_url": "", "post_date": "03/31/2020 21:49:05", "content": "2 final submissions selected. Good luck to everyone!
", "votes": null, "replies": [] } ] }, { "id": 792952, "author_name": "mpware", "author_url": "", "post_date": "03/31/2020 17:20:33", "content": "Yes, great journey! 2 last submissions in progress... The competition is over and I've learnt a lot once again. Kaggle is really a great place. Thanks to my motivated teammate @tikutiku that did great job with many models experiments and validation scheme totally different than mine that made our ensemble efficient (at least on public LB). Also, thanks to @timesler for his great facenet-pytorch MTCNN detector which we've used for most of our models. @timesler : Add TPU support and your lib will be the best π . Thanks to @humananalog that gives a boost to the competition by pushing us to do better with his kernel. Thanks to the organizers @juliaelliott for the TPU and GCP coupons. I've trained my first TPU (pytorch-XLA) model with this competition, it's really impressive with multi-cores, it's so fast.
\n\nThanks to all competitors for sharing idea and kernels. I can't wait to see solutions and approaches from top competitors. How can they reach 0.2? How did they deal with high quality fakes? How far their solution from State Of The Art? And finally how this competition will really help to solve the problem. If you look at https://www.thispersondoesnotexist.com/ it's a bit scary how great it is.
\n\nResults within the next weeks, shake up expected. βοΈ
", "votes": null, "replies": [] }, { "id": 793103, "author_name": "futurelyu", "author_url": "", "post_date": "03/31/2020 19:57:31", "content": "Yes, what a journey. This is also my first CV competition. Many thanks for the resources provided by AWS and GCP. Without their support, it's really hard to keep on this competition for a student who does not have great equipment.\nAnd I really appreciate the field of computer vision after the long learning journey. I have learnt a lot from the wonderful public notebook and discussions. Thanks for all competitors who share their ideas, especially @unkownhihi π .\nI can't wait to see the best solution and see whether it can really protect us from deepfake attack in real world!
", "votes": null, "replies": [] }, { "id": 793118, "author_name": "ankitsainiankit", "author_url": "", "post_date": "03/31/2020 20:15:11", "content": "I feel the same. It was a great journey. Everyone was helpful. Special thanks to @unkownhihi.
", "votes": null, "replies": [] }, { "id": 793152, "author_name": "", "author_url": "", "post_date": "03/31/2020 20:49:39", "content": "I really enjoyed the tournament. Yet, the competition was to real and I was only for learning :)
", "votes": null, "replies": [] }, { "id": 793177, "author_name": "kanyuenyong", "author_url": "", "post_date": "03/31/2020 21:17:54", "content": "I'm quite new here, but it's very exciting to join this platform. Thanks for sharing!
", "votes": null, "replies": [] }, { "id": 793241, "author_name": "purplejester", "author_url": "", "post_date": "03/31/2020 22:38:41", "content": "Indeed, amazing competition. Now I know that a dataset of ~450 GB of videos is not that scary as it seems :)
", "votes": null, "replies": [] }, { "id": 793305, "author_name": "unkownhihi", "author_url": "", "post_date": "03/31/2020 23:55:07", "content": "LOL Added more dramatic! We got our new score 10 minutes before deadline..... π¬
", "votes": null, "replies": [] }, { "id": 793473, "author_name": "maralski", "author_url": "", "post_date": "04/01/2020 03:54:43", "content": "Looking forward to my electricity bill (not!) π₯π₯
", "votes": null, "replies": [] }, { "id": 794370, "author_name": "larsen0966", "author_url": "", "post_date": "04/01/2020 19:04:58", "content": "Thank you for your strength and courage! our words help all of us in isolation to get through tough times.\nKeep up the good work!
", "votes": null, "replies": [] }, { "id": 794853, "author_name": "aktiwari5", "author_url": "", "post_date": "04/02/2020 06:25:58", "content": "Just getting started with this exciting journey towards DS, AI, DL ....
", "votes": null, "replies": [] }, { "id": 795562, "author_name": "sonalirangwani", "author_url": "", "post_date": "04/02/2020 20:34:22", "content": "Encouraging words!
", "votes": null, "replies": [] }, { "id": 796605, "author_name": "ruhilvikash1", "author_url": "", "post_date": "04/03/2020 18:28:07", "content": "Dear Fellow Kagglers,\nThis is my first comment here as I am starting my journey on Kaggle. I would appreciate all you guidance to make it through and work along withyou all.\nThanks
", "votes": null, "replies": [] }, { "id": 804856, "author_name": "rhedenjohn", "author_url": "", "post_date": "04/12/2020 03:42:02", "content": "Thank you for inspiring me, awesome story
", "votes": null, "replies": [] }, { "id": 806600, "author_name": "collinrm", "author_url": "", "post_date": "04/13/2020 21:44:20", "content": "Fantastic achievement. Truly inspiring.
", "votes": null, "replies": [] }, { "id": 807519, "author_name": "starzmustdie", "author_url": "", "post_date": "04/14/2020 17:38:44", "content": "congrats
", "votes": null, "replies": [] }, { "id": 812634, "author_name": "tariqjawed", "author_url": "", "post_date": "04/18/2020 21:37:03", "content": "encouraging words... π
", "votes": null, "replies": [] }, { "id": 818133, "author_name": "rashakdude", "author_url": "", "post_date": "04/23/2020 16:35:55", "content": "π
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "792416": "Hi all!\n\nGetting a bit emotional, and therefore I had to write.\n\nThis competition was a very special and emotional one for me on many levels. I am sure many other Kagglers, especially, who dared to take part in this competition in spite of having limited computing resources can relate as well! This is my first serious Kaggle competition, not an easy one to start for sure, but the amount of learning was so huge that I will never regret it.\n\nThanks to @unkownhihi and his team, for establishing themselves as public figures from early on who have selflessly helped so many Kagglers like me to start the competition with a simple public kernel, throwing away so many hints in discussions, and for continuing to share extremely helpful kernels. You are the best!\n\nThanks to @humananalog, for that controversial kernel, that in fact turned out to be one of the most helpful notebooks that gave away hints to many of us on how to successfully build a working dataset, train a transfer learning model to cross a bubble (that at some point of time I thought I will not be able to cross) around the log loss of 0.69LB. Kudos!\n\nThanks to everyone who contributed to the discussions- a place that literally becomes like a social media timeline, that I used to check when I am having my breakfast, dinner, lunch or when I am stuck in the subway. Literally, we all read it for entertainment.\n\nThanks to GCP, Amazon for timely cloud resources. Thanks to Kaggle for your super fast GPU (and UI modifications)! Without you, many of us would have been drifting without any directions.\n\nAnother thing that will make me never forget this competition is the unfortunate Covid-19 situation. We all suffered this way or that, but the competition kept going, showing science wins no matter what!\n\nIt all started in 2019, a year when social distancing was a concept that I never heard, and now writing it from my room, while I am self-isolated. No matter what happens next, as someone said, \"It's All About the Journey, Not the Destination\".\n\nIndeed, what a journey!\n\nThanks, everyone! All the best!", "792439": "May the force be with you ! I will keep my emotion for the potential big drop tomorrow π", "792442": "My final submission is running, so I donβt want any emotion or stress tomorrow :D Anyway, I believe it will take some time to run these submission in private set as this competition may be a special case. Also, they will give public score only for 2 submissions, thank god thatβs a relief from the panic for not submitting other potential good submissions that we donβt select.", "792469": "haha, you are right. It will take time to score, so at least I will cry another day :)", "792563": "Well put, I relate on many levels to what you said esp it being my first. Definitely a bittersweet end to a great competition.", "792902": "Yes, indeed, a great journey. Many ups and downs, surprises(good and bad), discoveries, new thoughts. Good luck at the private leaderboard to all!!!", "792904": "unknownhihi \nwhat do you think about the shakeup?", "792920": "Well, ensemble as many model as you can, do Kfold if possible. Stress test your model(but probably a little bit too late)\n\nWe, unfortunately, ran out of aws quota and had to terminate all the instances(which means permanently delete) that have our model on it. π£ Now the only thing we could do is cross our fingers.", "792943": ">> Well, ensemble as many model as you can, do Kfold if possible. Stress test your model(but probably a little bit too late)\n\nDone that, fingers crossed!", "792948": "Not sure I understood correctly your words about the big drop. There will be no drop tomorrow because the results will be revealed in the end of April.\n\n> (Estimated) April 22, 2020 - Private leaderboard revealed.\n\n>Note: Because the private test set is held outside of Kaggle's platform, the private leaderboard will >not be revealed immediately. We have estimated the scoring timeframe, but this is subject to revision.", "792952": "Yes, great journey! 2 last submissions in progress... The competition is over and I've learnt a lot once again. Kaggle is really a great place. Thanks to my motivated teammate @tikutiku that did great job with many models experiments and validation scheme totally different than mine that made our ensemble efficient (at least on public LB). Also, thanks to @timesler for his great facenet-pytorch MTCNN detector which we've used for most of our models. @timesler : Add TPU support and your lib will be the best π . Thanks to @humananalog that gives a boost to the competition by pushing us to do better with his kernel. Thanks to the organizers @juliaelliott for the TPU and GCP coupons. I've trained my first TPU (pytorch-XLA) model with this competition, it's really impressive with multi-cores, it's so fast.\n\nThanks to all competitors for sharing idea and kernels. I can't wait to see solutions and approaches from top competitors. How can they reach 0.2? How did they deal with high quality fakes? How far their solution from State Of The Art? And finally how this competition will really help to solve the problem. If you look at https://www.thispersondoesnotexist.com/ it's a bit scary how great it is.\n\nResults within the next weeks, shake up expected. βοΈ", "792975": "That's a long time. Time to focus on M5 :D", "793099": "That's what I was going to say as well. :) \nTime to focus on other competitions while top solutions are being scored. ;)\nBest of luck for all of you!", "793103": "Yes, what a journey. This is also my first CV competition. Many thanks for the resources provided by AWS and GCP. Without their support, it's really hard to keep on this competition for a student who does not have great equipment.\nAnd I really appreciate the field of computer vision after the long learning journey. I have learnt a lot from the wonderful public notebook and discussions. Thanks for all competitors who share their ideas, especially @unkownhihi π .\nI can't wait to see the best solution and see whether it can really protect us from deepfake attack in real world!", "793118": "I feel the same. It was a great journey. Everyone was helpful. Special thanks to @unkownhihi.", "793152": "I really enjoyed the tournament. Yet, the competition was to real and I was only for learning :)", "793177": "I'm quite new here, but it's very exciting to join this platform. Thanks for sharing!", "793195": "2 final submissions selected. Good luck to everyone!", "793241": "Indeed, amazing competition. Now I know that a dataset of ~450 GB of videos is not that scary as it seems :)", "793305": "LOL Added more dramatic! We got our new score 10 minutes before deadline..... π¬", "793473": "Looking forward to my electricity bill (not!) π₯π₯", "794370": "Thank you for your strength and courage! our words help all of us in isolation to get through tough times.\nKeep up the good work!", "794853": "Just getting started with this exciting journey towards DS, AI, DL ....", "795562": "Encouraging words!", "796605": "Dear Fellow Kagglers,\nThis is my first comment here as I am starting my journey on Kaggle. I would appreciate all you guidance to make it through and work along withyou all.\nThanks", "804856": "Thank you for inspiring me, awesome story", "806600": "Fantastic achievement. Truly inspiring.", "807519": "congrats", "812634": "encouraging words... π", "818133": "π" }, "source": "meta" }