![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3033466%2F93814776c87551a0e42c667cff54416d%2Fdogs.png?generation=1565847995479444&alt=media\")
Because the existing LB doesn't work anymore and The competition details say '17 days to go',maybe the Officials need to spend some time revising rankings. but i and my teammates can't wait to know which medal area we are in. so, so Let's share the real LB score?
\n\nLet me first, we achieved 24.95 using DCGAN. I think if we were given more time, we would do better. Finally, thank my teammates for all their efforts.
", "messages": [ { "id": 599551, "postDate": "2019-08-15T05:33:56.247Z", "content": "Because the existing LB doesn't work anymore and The competition details say '17 days to go',maybe the Officials need to spend some time revising rankings. but i and my teammates can't wait to know which medal area we are in. so, so Let's share the real LB score?
\n\nLet me first, we achieved 24.95 using DCGAN. I think if we were given more time, we would do better. Finally, thank my teammates for all their efforts.
", "rawMarkdown": "Because the existing LB doesn't work anymore and The competition details say '17 days to go',maybe the Officials need to spend some time revising rankings. but i and my teammates can't wait to know which medal area we are in. so, so Let's share the real LB score?\n\nLet me first, we achieved 24.95 using DCGAN. I think if we were given more time, we would do better. Finally, thank my teammates for all their efforts.", "votes": 9 }, { "id": 599946, "postDate": "2019-08-15T13:30:44.883Z", "content": "1.bestfitting [12.46326]
\n\n2.Dmitry Vorobiev[14.95516]
\n\n3.yabea [15.81605]
\n\n4.X5(for cheating)[24.95]
\n\nTheo Viel [30.33340]
\n\nAlvaro[30.65]
\n\nTom[31.06344]
\n\nNaomichi Agata [36.76123]
\n\nJohannl[38.26191]
\n\nLeon Shangguan [38.70295]
\n\nHappy GAN[38.70295]
\n\nhirune924[38.92350]
\n\nNanash [30-40]
\n\nFunkyboy [44.69078].
\n\nfujibo[45.6]
\n\nTimetraveler [47.02282]
\n\nChris Deotte [47.28169].
\n\nFelipe Loque [47.51796]
\n\nMomi64 [47.75757].
\n\nMighty Rains [54.02973].
\n\n\"Young for you\" [54.11119].
", "rawMarkdown": "1.bestfitting [12.46326]\n\n2.Dmitry Vorobiev[14.95516] \n\n3.yabea [15.81605]\n\n4.X5(for cheating)[24.95]\n\nTheo Viel [30.33340]\n\nAlvaro[30.65]\n\nTom[31.06344]\n\nNaomichi Agata [36.76123]\n\nJohannl[38.26191] \n\n Leon Shangguan [38.70295]\n\nHappy GAN[38.70295] \n\nhirune924[38.92350]\n\nNanash [30-40] \n\nFunkyboy [44.69078].\n\nfujibo[45.6]\n\nTimetraveler [47.02282]\n\nChris Deotte [47.28169].\n\nFelipe Loque [47.51796]\n\n Momi64 [47.75757].\n\nMighty Rains [54.02973].\n\n\"Young for you\" [54.11119].", "votes": 8, "replies": [ { "id": 600260, "postDate": "2019-08-15T21:10:28.333Z", "content": "You forgot to add your team at [24.95]. Also Bestfitting is at [12.46326] and Timetraveler is [47.02282]. Then here are all the posted kernels under 55. Alvaro at [30.65]. Leon Shangguan at [38.70295]. Funkyboy at [44.69078]. Chris Deotte at [47.28169]. Felipe Loque at [47.51796]. Momi64 at [47.75757]. Mighty Rains at [54.02973]. And \"Young for you\" at [54.11119].
\n\nAnd then here's a list of others based on downloading the public leaderboard. But there's probably 50 more under LB 55 that we don't know about.
", "rawMarkdown": "You forgot to add your team at [24.95]. Also Bestfitting is at [12.46326] and Timetraveler is [47.02282]. Then here are all the posted kernels under 55. Alvaro at [30.65]. Leon Shangguan at [38.70295]. Funkyboy at [44.69078]. Chris Deotte at [47.28169]. Felipe Loque at [47.51796]. Momi64 at [47.75757]. Mighty Rains at [54.02973]. And \"Young for you\" at [54.11119].\n\nAnd then [here's][1] a list of others based on downloading the public leaderboard. But there's probably 50 more under LB 55 that we don't know about.\n\n[1]: https://www.kaggle.com/c/generative-dog-images/discussion/104028", "votes": 4 }, { "id": 600349, "postDate": "2019-08-16T02:28:36.360Z", "content": "thx chirs ,i will update it
", "rawMarkdown": "thx chirs ,i will update it " } ] }, { "id": 599556, "postDate": "2019-08-15T05:47:01.757Z", "content": "The dog images\n
look great
", "rawMarkdown": "look great" }, { "id": 599983, "postDate": "2019-08-15T13:54:26.123Z", "content": "thanks chris, It feels great to compete with you.（always learn many things from you）thx❤️
", "rawMarkdown": "thanks chris, It feels great to compete with you.（always learn many things from you）thx❤️ ", "votes": 2 } ] }, { "id": 599663, "postDate": "2019-08-15T08:30:12.723Z", "content": "I am at 47\nI don't even expect a silver, will probably.. maybe.. get a bronze.
", "rawMarkdown": "I am at 47\nI don't even expect a silver, will probably.. maybe.. get a bronze. ", "votes": 1, "replies": [ { "id": 599818, "postDate": "2019-08-15T11:29:08.427Z", "content": "thanks for sharing ❤️
", "rawMarkdown": "thanks for sharing ❤️ " } ] }, { "id": 599606, "postDate": "2019-08-15T07:08:45.113Z", "content": "wow 24.95 is really impressive! amazing work :) I'm probably between 30-40 👍
", "rawMarkdown": "wow 24.95 is really impressive! amazing work :) I'm probably between 30-40 👍 ", "votes": 1, "replies": [ { "id": 599609, "postDate": "2019-08-15T07:10:19.023Z", "content": "thanks for sharing Nanashi❤️
", "rawMarkdown": "thanks for sharing Nanashi❤️ " } ] }, { "id": 599872, "postDate": "2019-08-15T12:13:04.703Z", "content": "LB 24.95 with a DCGAN is great. What features did you use? Was it a conditional GAN, what normalizations did you use, what loss was used, etc.
", "rawMarkdown": "LB 24.95 with a DCGAN is great. What features did you use? Was it a conditional GAN, what normalizations did you use, what loss was used, etc.", "replies": [ { "id": 599941, "postDate": "2019-08-15T13:26:55.403Z", "content": "umm... We may make our code open if at the gold area. Shortly, \n- data preprocess: 4 crops for one bbox.\n- cudnn benchmark = True for speeding up in Pytorch.\n- forward twice and backward twice per iteration for speeding up. For example, change D(real) & D(fake) to D(cat[real, fake]), but BN should be computed separately, so we define a Split BN op.\n- go deeper and wider (especially for G).\n- label smooth + Cross Entropy w/ margin for preventing D from overfiting.\n- EMA w/ beta increasing gradually & fixed lr.\nNo conditional. G and D use BN and Split BN respectively. Very thanks for your contribute that is a great inspiration to us. 👍
", "rawMarkdown": "umm... We may make our code open if at the gold area. Shortly, \n- data preprocess: 4 crops for one bbox.\n- cudnn benchmark = True for speeding up in Pytorch.\n- forward twice and backward twice per iteration for speeding up. For example, change D(real) & D(fake) to D(cat[real, fake]), but BN should be computed separately, so we define a Split BN op.\n- go deeper and wider (especially for G).\n- label smooth + Cross Entropy w/ margin for preventing D from overfiting.\n- EMA w/ beta increasing gradually & fixed lr.\nNo conditional. G and D use BN and Split BN respectively. Very thanks for your contribute that is a great inspiration to us. 👍 ", "votes": 4 } ] }, { "id": 599555, "postDate": "2019-08-15T05:46:08.957Z", "content": "Oh, my smart teammate used a God's image crop method and memorized GAN and got 1st place. 💔
", "rawMarkdown": "Oh, my smart teammate used a God's image crop method and memorized GAN and got 1st place. 💔 ", "replies": [ { "id": 599559, "postDate": "2019-08-15T05:53:41.133Z", "content": "There is no doubt that memorized GAN will be forbidden.a crop method with some little trick can get the top. 😂
", "rawMarkdown": "There is no doubt that memorized GAN will be forbidden.a crop method with some little trick can get the top. 😂 " } ] }, { "id": 700235, "postDate": "2019-12-21T17:04:34.390Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 599946, "author_name": "MeisterMorxrc", "author_url": "", "post_date": "2019-08-15T13:30:44.883000", "content": "1.bestfitting [12.46326]
\n\n2.Dmitry Vorobiev[14.95516]
\n\n3.yabea [15.81605]
\n\n4.X5(for cheating)[24.95]
\n\nTheo Viel [30.33340]
\n\nAlvaro[30.65]
\n\nTom[31.06344]
\n\nNaomichi Agata [36.76123]
\n\nJohannl[38.26191]
\n\nLeon Shangguan [38.70295]
\n\nHappy GAN[38.70295]
\n\nhirune924[38.92350]
\n\nNanash [30-40]
\n\nFunkyboy [44.69078].
\n\nfujibo[45.6]
\n\nTimetraveler [47.02282]
\n\nChris Deotte [47.28169].
\n\nFelipe Loque [47.51796]
\n\nMomi64 [47.75757].
\n\nMighty Rains [54.02973].
\n\n\"Young for you\" [54.11119].
", "votes": 8, "replies": [ { "id": 600260, "author_name": "Chris Deotte", "author_url": "", "post_date": "2019-08-15T21:10:28.333000", "content": "You forgot to add your team at [24.95]. Also Bestfitting is at [12.46326] and Timetraveler is [47.02282]. Then here are all the posted kernels under 55. Alvaro at [30.65]. Leon Shangguan at [38.70295]. Funkyboy at [44.69078]. Chris Deotte at [47.28169]. Felipe Loque at [47.51796]. Momi64 at [47.75757]. Mighty Rains at [54.02973]. And \"Young for you\" at [54.11119].
\n\nAnd then here's a list of others based on downloading the public leaderboard. But there's probably 50 more under LB 55 that we don't know about.
", "votes": 4, "replies": [] }, { "id": 600349, "author_name": "MeisterMorxrc", "author_url": "", "post_date": "2019-08-16T02:28:36.360000", "content": "thx chirs ,i will update it
", "votes": 0, "replies": [] } ] }, { "id": 599556, "author_name": "MeisterMorxrc", "author_url": "", "post_date": "2019-08-15T05:47:01.757000", "content": "The dog images\n
look great
", "votes": 0, "replies": [] }, { "id": 599983, "author_name": "MeisterMorxrc", "author_url": "", "post_date": "2019-08-15T13:54:26.123000", "content": "thanks chris, It feels great to compete with you.（always learn many things from you）thx❤️
", "votes": 2, "replies": [] } ] }, { "id": 599663, "author_name": "timetraveller", "author_url": "", "post_date": "2019-08-15T08:30:12.723000", "content": "I am at 47\nI don't even expect a silver, will probably.. maybe.. get a bronze.
", "votes": 1, "replies": [ { "id": 599818, "author_name": "MeisterMorxrc", "author_url": "", "post_date": "2019-08-15T11:29:08.427000", "content": "thanks for sharing ❤️
", "votes": 0, "replies": [] } ] }, { "id": 599606, "author_name": "Nanashi", "author_url": "", "post_date": "2019-08-15T07:08:45.113000", "content": "wow 24.95 is really impressive! amazing work :) I'm probably between 30-40 👍
", "votes": 1, "replies": [ { "id": 599609, "author_name": "MeisterMorxrc", "author_url": "", "post_date": "2019-08-15T07:10:19.023000", "content": "thanks for sharing Nanashi❤️
", "votes": 0, "replies": [] } ] }, { "id": 599872, "author_name": "Chris Deotte", "author_url": "", "post_date": "2019-08-15T12:13:04.703000", "content": "LB 24.95 with a DCGAN is great. What features did you use? Was it a conditional GAN, what normalizations did you use, what loss was used, etc.
", "votes": 0, "replies": [ { "id": 599941, "author_name": "baigou", "author_url": "", "post_date": "2019-08-15T13:26:55.403000", "content": "umm... We may make our code open if at the gold area. Shortly, \n- data preprocess: 4 crops for one bbox.\n- cudnn benchmark = True for speeding up in Pytorch.\n- forward twice and backward twice per iteration for speeding up. For example, change D(real) & D(fake) to D(cat[real, fake]), but BN should be computed separately, so we define a Split BN op.\n- go deeper and wider (especially for G).\n- label smooth + Cross Entropy w/ margin for preventing D from overfiting.\n- EMA w/ beta increasing gradually & fixed lr.\nNo conditional. G and D use BN and Split BN respectively. Very thanks for your contribute that is a great inspiration to us. 👍
", "votes": 4, "replies": [] } ] }, { "id": 599555, "author_name": "baigou", "author_url": "", "post_date": "2019-08-15T05:46:08.957000", "content": "Oh, my smart teammate used a God's image crop method and memorized GAN and got 1st place. 💔
", "votes": 0, "replies": [ { "id": 599559, "author_name": "MeisterMorxrc", "author_url": "", "post_date": "2019-08-15T05:53:41.133000", "content": "There is no doubt that memorized GAN will be forbidden.a crop method with some little trick can get the top. 😂
", "votes": 0, "replies": [] } ] }, { "id": 700235, "author_name": "", "author_url": "", "post_date": "2019-12-21T17:04:34.390000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "599551": "Because the existing LB doesn't work anymore and The competition details say '17 days to go',maybe the Officials need to spend some time revising rankings. but i and my teammates can't wait to know which medal area we are in. so, so Let's share the real LB score?\n\nLet me first, we achieved 24.95 using DCGAN. I think if we were given more time, we would do better. Finally, thank my teammates for all their efforts.", "599946": "1.bestfitting [12.46326]\n\n2.Dmitry Vorobiev[14.95516] \n\n3.yabea [15.81605]\n\n4.X5(for cheating)[24.95]\n\nTheo Viel [30.33340]\n\nAlvaro[30.65]\n\nTom[31.06344]\n\nNaomichi Agata [36.76123]\n\nJohannl[38.26191] \n\n Leon Shangguan [38.70295]\n\nHappy GAN[38.70295] \n\nhirune924[38.92350]\n\nNanash [30-40] \n\nFunkyboy [44.69078].\n\nfujibo[45.6]\n\nTimetraveler [47.02282]\n\nChris Deotte [47.28169].\n\nFelipe Loque [47.51796]\n\n Momi64 [47.75757].\n\nMighty Rains [54.02973].\n\n\"Young for you\" [54.11119].", "599556": "The dog images\n\n", "599663": "I am at 47\nI don't even expect a silver, will probably.. maybe.. get a bronze. ", "599606": "wow 24.95 is really impressive! amazing work :) I'm probably between 30-40 👍 ", "599872": "LB 24.95 with a DCGAN is great. What features did you use? Was it a conditional GAN, what normalizations did you use, what loss was used, etc.", "599555": "Oh, my smart teammate used a God's image crop method and memorized GAN and got 1st place. 💔 ", "700235": "" } }