Thank you for your participation in this competition. The private leaderboard has now been published, and rankings are finalized. Congratulations to the winners!
\n\nPlease note the following:
\n\n999999. We reviewed more than 700 Kernels, in order to make sure everyone who gets a medal has a legitimate submission. Common prohibited methods that we saw were:\n```cosine_distance_eps in the evaluation metric (https://www.kaggle.com/wendykan/demo-mifid-metric-for-dog-image-generation-comp) for the memorization distance in the private model’s latent space in order to properly penalize the memorization GANs. We are so impressed with all of the enthusiasm and engagement around this competition. Thank you for your patience and tolerance during this experimental competition format. We hope to continue to bring you advanced competition types in the future!
", "messages": [ { "id": "610419", "postDate": "08/28/2019 19:28:06", "content": "Thank you for your participation in this competition. The private leaderboard has now been published, and rankings are finalized. Congratulations to the winners!
\n\nPlease note the following:
\n\n999999. We reviewed more than 700 Kernels, in order to make sure everyone who gets a medal has a legitimate submission. Common prohibited methods that we saw were:\n```cosine_distance_eps in the evaluation metric (https://www.kaggle.com/wendykan/demo-mifid-metric-for-dog-image-generation-comp) for the memorization distance in the private model’s latent space in order to properly penalize the memorization GANs. We are so impressed with all of the enthusiasm and engagement around this competition. Thank you for your patience and tolerance during this experimental competition format. We hope to continue to bring you advanced competition types in the future!
", "rawMarkdown": "Thank you for your participation in this competition. The private leaderboard has now been published, and rankings are finalized. Congratulations to the winners!\n\nPlease note the following:\n\n- Submissions which used prohibited methods were manually inspected and invalidated by setting private scores to `999999`. We reviewed more than 700 Kernels, in order to make sure everyone who gets a medal has a legitimate submission. Common prohibited methods that we saw were:\n```\n- Memorization GANs: Training a discriminator to fit the training set, freezing the discriminator, and then training the generator\n- Mapping 10k-one-hot vectors to the 10k training images\n- VAE Reconstruction: Training a VAE, but generating results by reconstructing training images\n- Augmentation or blending (cropping, morphing, adding noise, etc.) of training set\n```\n- As we did the manual inspection of top memorization GANs, we adjusted the memorization threshold `cosine_distance_eps` in the evaluation metric (https://www.kaggle.com/wendykan/demo-mifid-metric-for-dog-image-generation-comp) for the memorization distance in the private model’s latent space in order to properly penalize the memorization GANs. \n- Submission that failed to run were invalidated.\n- Submissions with deleted kernels were invalidated. \n- We made sure everyone’s two selected (or auto-selected) kernels were rescored. The private leaderboard rankings are based only on selected/auto-selected kernels. We also tried to provide scores on your other submissions to give you knowledge on how they performed, but non-selected submissions that failed to run were invalidated. \n- Note that late submissions will not be enabled for this competition.\n- If you feel your submission is based on a non-prohibited generative model and therefore wrongfully invalidated, you may submit a claim through our [compliance channel](https://www.kaggle.com/compliance).\n________________\nWe are so impressed with all of the enthusiasm and engagement around this competition. Thank you for your patience and tolerance during this experimental competition format. We hope to continue to bring you advanced competition types in the future!", "votes": null }, { "id": "610422", "postDate": "08/28/2019 19:29:58", "content": "Well done all! Including to the orgsnisers, who were thoughtful and receptive.
\n\nWhat was the private dataset? I was expecting it to be Columbia dogs but had no time to tune results against it.
", "rawMarkdown": "Well done all! Including to the orgsnisers, who were thoughtful and receptive.\n\nWhat was the private dataset? I was expecting it to be Columbia dogs but had no time to tune results against it.", "votes": null }, { "id": "610547", "postDate": "08/28/2019 21:21:38", "content": "The first bronze model :)
\n\nThough my solution was super-simple, I've learned a lot about GANs. The winners' solutions are much more efficient, of course, and now it is time to look into them thoroughly.
", "rawMarkdown": "The first bronze model :)\n\nThough my solution was super-simple, I've learned a lot about GANs. The winners' solutions are much more efficient, of course, and now it is time to look into them thoroughly.", "votes": null }, { "id": "610579", "postDate": "08/28/2019 22:01:36", "content": "It's a private dataset with a mix of collected dog images + Kaggle employees' dogs :)
", "rawMarkdown": "It's a private dataset with a mix of collected dog images + Kaggle employees' dogs :)", "votes": null }, { "id": "610723", "postDate": "08/29/2019 00:36:03", "content": "@wendykan. It was a great learning experience for me in this competition. Looking forward to many more such competitions. Thanks to the organizers and the kaggle team for conducting these.
", "rawMarkdown": "wendykan. It was a great learning experience for me in this competition. Looking forward to many more such competitions. Thanks to the organizers and the kaggle team for conducting these.", "votes": null }, { "id": "610937", "postDate": "08/29/2019 03:11:54", "content": "Thanks for being so patient to review these kernels. It's an excellent and unique competition and I have learned a lot about GANs. I'm looking forward to more competitions about generative models.
", "rawMarkdown": "Thanks for being so patient to review these kernels. It's an excellent and unique competition and I have learned a lot about GANs. I'm looking forward to more competitions about generative models.", "votes": null }, { "id": "611004", "postDate": "08/29/2019 03:59:10", "content": "Thanks to the organizers and the kaggle team for conducting these. Thanks for being so patient to review these kernels. It's an excellent and unique competition and I have learned a lot about GANs. I'm looking forward to more competitions about generative models... @wendykan
", "rawMarkdown": "Thanks to the organizers and the kaggle team for conducting these. Thanks for being so patient to review these kernels. It's an excellent and unique competition and I have learned a lot about GANs. I'm looking forward to more competitions about generative models... @wendykan", "votes": null }, { "id": "611070", "postDate": "08/29/2019 04:49:24", "content": "@wendykan @andrewcybai @juliaelliott Thanks a lot to your huge efforts and support in organizing this competition and reviewing most of the kernel submissions. I'm glad to be part of this competition and really learnt a lot about GANs through it! Innovative setup for GAN competition as well, great job!
", "rawMarkdown": "wendykan @andrewcybai @juliaelliott Thanks a lot to your huge efforts and support in organizing this competition and reviewing most of the kernel submissions. I'm glad to be part of this competition and really learnt a lot about GANs through it! Innovative setup for GAN competition as well, great job!", "votes": null }, { "id": "611158", "postDate": "08/29/2019 06:22:44", "content": "Thank you very much for your efforts throughout the competition. Most notably, your solutions to memorizer GANs and extension of the competition. Looking forward to more generative model competitions.
", "rawMarkdown": "Thank you very much for your efforts throughout the competition. Most notably, your solutions to memorizer GANs and extension of the competition. Looking forward to more generative model competitions.", "votes": null }, { "id": "611277", "postDate": "08/29/2019 07:41:56", "content": "@wendykan ...thank you team Kaggle for hosting this competition. When I started this competition, I had limited knowledge of GANs but I learned about it and participate in this competition. Now, I had good knowledge about GANs.
\n\nIt was great learning experience for me and I enjoy a lot by participating in this competition.
\n\nAlso, thank you organizers for organizing this competition.
", "rawMarkdown": "wendykan ...thank you team Kaggle for hosting this competition. When I started this competition, I had limited knowledge of GANs but I learned about it and participate in this competition. Now, I had good knowledge about GANs.\n\nIt was great learning experience for me and I enjoy a lot by participating in this competition.\n\nAlso, thank you organizers for organizing this competition.", "votes": null }, { "id": "611403", "postDate": "08/29/2019 08:42:49", "content": "Thank you, Wendy.
", "rawMarkdown": "Thank you, Wendy.", "votes": null }, { "id": "611602", "postDate": "08/29/2019 10:44:26", "content": "Thank you for organizing this comptition. \nI had a lot of fun competing and learned a lot of new things.
", "rawMarkdown": "Thank you for organizing this comptition. \nI had a lot of fun competing and learned a lot of new things.", "votes": null }, { "id": "611770", "postDate": "08/29/2019 12:26:47", "content": "Thanks! It was a very fun competition and a very distinct experience from all other DL staff I've tried before
", "rawMarkdown": "Thanks! It was a very fun competition and a very distinct experience from all other DL staff I've tried before", "votes": null }, { "id": "612257", "postDate": "08/29/2019 17:23:25", "content": "Thanks Wendy. During this competition we had a lot of mess due its experimental format but I think you did a huge work to inspect all the notebooks and provide reasonable results! Will be waiting for more GANs competitions.
", "rawMarkdown": "Thanks Wendy. During this competition we had a lot of mess due its experimental format but I think you did a huge work to inspect all the notebooks and provide reasonable results! Will be waiting for more GANs competitions.", "votes": null } ], "comments": [ { "id": 610422, "author_name": "robga", "author_url": "", "post_date": "08/28/2019 19:29:58", "content": "Well done all! Including to the orgsnisers, who were thoughtful and receptive.
\n\nWhat was the private dataset? I was expecting it to be Columbia dogs but had no time to tune results against it.
", "votes": null, "replies": [ { "id": 610579, "author_name": "wendykan", "author_url": "", "post_date": "08/28/2019 22:01:36", "content": "It's a private dataset with a mix of collected dog images + Kaggle employees' dogs :)
", "votes": null, "replies": [] } ] }, { "id": 610547, "author_name": "purplejester", "author_url": "", "post_date": "08/28/2019 21:21:38", "content": "The first bronze model :)
\n\nThough my solution was super-simple, I've learned a lot about GANs. The winners' solutions are much more efficient, of course, and now it is time to look into them thoroughly.
", "votes": null, "replies": [] }, { "id": 610723, "author_name": "manojprabhaakr", "author_url": "", "post_date": "08/29/2019 00:36:03", "content": "@wendykan. It was a great learning experience for me in this competition. Looking forward to many more such competitions. Thanks to the organizers and the kaggle team for conducting these.
", "votes": null, "replies": [] }, { "id": 610937, "author_name": "lyonhou", "author_url": "", "post_date": "08/29/2019 03:11:54", "content": "Thanks for being so patient to review these kernels. It's an excellent and unique competition and I have learned a lot about GANs. I'm looking forward to more competitions about generative models.
", "votes": null, "replies": [] }, { "id": 611004, "author_name": "veeralakrishna", "author_url": "", "post_date": "08/29/2019 03:59:10", "content": "Thanks to the organizers and the kaggle team for conducting these. Thanks for being so patient to review these kernels. It's an excellent and unique competition and I have learned a lot about GANs. I'm looking forward to more competitions about generative models... @wendykan
", "votes": null, "replies": [] }, { "id": 611070, "author_name": "markpeng", "author_url": "", "post_date": "08/29/2019 04:49:24", "content": "@wendykan @andrewcybai @juliaelliott Thanks a lot to your huge efforts and support in organizing this competition and reviewing most of the kernel submissions. I'm glad to be part of this competition and really learnt a lot about GANs through it! Innovative setup for GAN competition as well, great job!
", "votes": null, "replies": [] }, { "id": 611158, "author_name": "shaygu", "author_url": "", "post_date": "08/29/2019 06:22:44", "content": "Thank you very much for your efforts throughout the competition. Most notably, your solutions to memorizer GANs and extension of the competition. Looking forward to more generative model competitions.
", "votes": null, "replies": [] }, { "id": 611277, "author_name": "prashant111", "author_url": "", "post_date": "08/29/2019 07:41:56", "content": "@wendykan ...thank you team Kaggle for hosting this competition. When I started this competition, I had limited knowledge of GANs but I learned about it and participate in this competition. Now, I had good knowledge about GANs.
\n\nIt was great learning experience for me and I enjoy a lot by participating in this competition.
\n\nAlso, thank you organizers for organizing this competition.
", "votes": null, "replies": [] }, { "id": 611403, "author_name": "grapestone5321", "author_url": "", "post_date": "08/29/2019 08:42:49", "content": "Thank you, Wendy.
", "votes": null, "replies": [] }, { "id": 611602, "author_name": "timetraveller98", "author_url": "", "post_date": "08/29/2019 10:44:26", "content": "Thank you for organizing this comptition. \nI had a lot of fun competing and learned a lot of new things.
", "votes": null, "replies": [] }, { "id": 611770, "author_name": "dvorobiev", "author_url": "", "post_date": "08/29/2019 12:26:47", "content": "Thanks! It was a very fun competition and a very distinct experience from all other DL staff I've tried before
", "votes": null, "replies": [] }, { "id": 612257, "author_name": "demonplus", "author_url": "", "post_date": "08/29/2019 17:23:25", "content": "Thanks Wendy. During this competition we had a lot of mess due its experimental format but I think you did a huge work to inspect all the notebooks and provide reasonable results! Will be waiting for more GANs competitions.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "610419": "Thank you for your participation in this competition. The private leaderboard has now been published, and rankings are finalized. Congratulations to the winners!\n\nPlease note the following:\n\n- Submissions which used prohibited methods were manually inspected and invalidated by setting private scores to `999999`. We reviewed more than 700 Kernels, in order to make sure everyone who gets a medal has a legitimate submission. Common prohibited methods that we saw were:\n```\n- Memorization GANs: Training a discriminator to fit the training set, freezing the discriminator, and then training the generator\n- Mapping 10k-one-hot vectors to the 10k training images\n- VAE Reconstruction: Training a VAE, but generating results by reconstructing training images\n- Augmentation or blending (cropping, morphing, adding noise, etc.) of training set\n```\n- As we did the manual inspection of top memorization GANs, we adjusted the memorization threshold `cosine_distance_eps` in the evaluation metric (https://www.kaggle.com/wendykan/demo-mifid-metric-for-dog-image-generation-comp) for the memorization distance in the private model’s latent space in order to properly penalize the memorization GANs. \n- Submission that failed to run were invalidated.\n- Submissions with deleted kernels were invalidated. \n- We made sure everyone’s two selected (or auto-selected) kernels were rescored. The private leaderboard rankings are based only on selected/auto-selected kernels. We also tried to provide scores on your other submissions to give you knowledge on how they performed, but non-selected submissions that failed to run were invalidated. \n- Note that late submissions will not be enabled for this competition.\n- If you feel your submission is based on a non-prohibited generative model and therefore wrongfully invalidated, you may submit a claim through our [compliance channel](https://www.kaggle.com/compliance).\n________________\nWe are so impressed with all of the enthusiasm and engagement around this competition. Thank you for your patience and tolerance during this experimental competition format. We hope to continue to bring you advanced competition types in the future!", "610422": "Well done all! Including to the orgsnisers, who were thoughtful and receptive.\n\nWhat was the private dataset? I was expecting it to be Columbia dogs but had no time to tune results against it.", "610547": "The first bronze model :)\n\nThough my solution was super-simple, I've learned a lot about GANs. The winners' solutions are much more efficient, of course, and now it is time to look into them thoroughly.", "610579": "It's a private dataset with a mix of collected dog images + Kaggle employees' dogs :)", "610723": "wendykan. It was a great learning experience for me in this competition. Looking forward to many more such competitions. Thanks to the organizers and the kaggle team for conducting these.", "610937": "Thanks for being so patient to review these kernels. It's an excellent and unique competition and I have learned a lot about GANs. I'm looking forward to more competitions about generative models.", "611004": "Thanks to the organizers and the kaggle team for conducting these. Thanks for being so patient to review these kernels. It's an excellent and unique competition and I have learned a lot about GANs. I'm looking forward to more competitions about generative models... @wendykan", "611070": "wendykan @andrewcybai @juliaelliott Thanks a lot to your huge efforts and support in organizing this competition and reviewing most of the kernel submissions. I'm glad to be part of this competition and really learnt a lot about GANs through it! Innovative setup for GAN competition as well, great job!", "611158": "Thank you very much for your efforts throughout the competition. Most notably, your solutions to memorizer GANs and extension of the competition. Looking forward to more generative model competitions.", "611277": "wendykan ...thank you team Kaggle for hosting this competition. When I started this competition, I had limited knowledge of GANs but I learned about it and participate in this competition. Now, I had good knowledge about GANs.\n\nIt was great learning experience for me and I enjoy a lot by participating in this competition.\n\nAlso, thank you organizers for organizing this competition.", "611403": "Thank you, Wendy.", "611602": "Thank you for organizing this comptition. \nI had a lot of fun competing and learned a lot of new things.", "611770": "Thanks! It was a very fun competition and a very distinct experience from all other DL staff I've tried before", "612257": "Thanks Wendy. During this competition we had a lot of mess due its experimental format but I think you did a huge work to inspect all the notebooks and provide reasonable results! Will be waiting for more GANs competitions." }, "source": "meta" }