And that by using a singel DELG model+PP without ensemble, finally used the GLDv2 version to the baseline notebook, even using the GLDv1+PP resulted in a medal score.
\nAnd congrats to the winners, what an achievement!
\nIt will be valuable reading the write-ups.
And that by using a singel DELG model+PP without ensemble, finally used the GLDv2 version to the baseline notebook, even using the GLDv1+PP resulted in a medal score.
\nAnd congrats to the winners, what an achievement!
\nIt will be valuable reading the write-ups.
Congrats!
", "rawMarkdown": "Congrats!", "votes": null }, { "id": "1032188", "postDate": "09/30/2020 03:18:51", "content": "Wow, Congrats ! U r probably the highest ranked Expert I have seen 😄
", "rawMarkdown": "Wow, Congrats ! U r probably the highest ranked Expert I have seen 😄", "votes": null }, { "id": "1032230", "postDate": "09/30/2020 03:52:43", "content": "Congrats! May I ask you how you got such score just by modifying RANSAC parameters?
", "rawMarkdown": "Congrats! May I ask you how you got such score just by modifying RANSAC parameters?", "votes": null }, { "id": "1032436", "postDate": "09/30/2020 07:48:00", "content": "Congratulations!
", "rawMarkdown": "Congratulations!", "votes": null }, { "id": "1032443", "postDate": "09/30/2020 07:53:49", "content": "Thanks, and same for you! Yes that was one part, RANSAC parameters, then also some other pre/post-processing, and used only the orginal pre-trained weight from the DELG-site, no extra training. Will make a write-up/summary later. Did though trained a couple of EffNets 4-7, but the limit of 40h/week on Kaggle TPU and 6h/day on Colab TPU, I should started the training much earlier in the competition to use them in final ensemble.
", "rawMarkdown": "Thanks, and same for you! Yes that was one part, RANSAC parameters, then also some other pre/post-processing, and used only the orginal pre-trained weight from the DELG-site, no extra training. Will make a write-up/summary later. Did though trained a couple of EffNets 4-7, but the limit of 40h/week on Kaggle TPU and 6h/day on Colab TPU, I should started the training much earlier in the competition to use them in final ensemble.", "votes": null }, { "id": "1032454", "postDate": "09/30/2020 08:02:58", "content": "Thanks! Haha maybe can be the case, and will be even higher after this competition, many bronze medals and many many competitions just some points below the medalzone but still got ranking points, and also competitions with only ranking points as final prizes, well well, one have to be satisfied with that ranking after only one year with Kaggle :) Expert on being an Expert! ;)
", "rawMarkdown": "Thanks! Haha maybe can be the case, and will be even higher after this competition, many bronze medals and many many competitions just some points below the medalzone but still got ranking points, and also competitions with only ranking points as final prizes, well well, one have to be satisfied with that ranking after only one year with Kaggle :) Expert on being an Expert! ;)", "votes": null }, { "id": "1032456", "postDate": "09/30/2020 08:03:35", "content": "Thanks, and same for you! :)
", "rawMarkdown": "Thanks, and same for you! :)", "votes": null }, { "id": "1033006", "postDate": "09/30/2020 15:37:56", "content": "Wow, great work 😍
\nCongratulations @kirderf on your first silver ✌️
Thanks! :) Yes hard work and lots of reading, it paid off. Now with 1y with Kaggle, with almost different challenges/problems every time I hope some will return my 2nd year, to face the task even better the second time with some experience :)
", "rawMarkdown": "Thanks! :) Yes hard work and lots of reading, it paid off. Now with 1y with Kaggle, with almost different challenges/problems every time I hope some will return my 2nd year, to face the task even better the second time with some experience :)", "votes": null }, { "id": "1033754", "postDate": "10/01/2020 08:40:20", "content": "Yes, if you keep trying and learning, one day it'll certainly pay off.
\nSomeday, we may be able to win a Gold ✌️
Congratulations!
", "rawMarkdown": "Congratulations!", "votes": null }, { "id": "1040906", "postDate": "10/07/2020 12:41:31", "content": "Congratulation! all I did so far is filling profile😅 your achievement seems really great😃
", "rawMarkdown": "Congratulation! all I did so far is filling profile😅 your achievement seems really great😃", "votes": null } ], "comments": [ { "id": 1032171, "author_name": "tangshuyun", "author_url": "", "post_date": "09/30/2020 02:41:47", "content": "Congrats!
", "votes": null, "replies": [ { "id": 1032456, "author_name": "kirderf", "author_url": "", "post_date": "09/30/2020 08:03:35", "content": "Thanks, and same for you! :)
", "votes": null, "replies": [] } ] }, { "id": 1032188, "author_name": "sidneyng", "author_url": "", "post_date": "09/30/2020 03:18:51", "content": "Wow, Congrats ! U r probably the highest ranked Expert I have seen 😄
", "votes": null, "replies": [ { "id": 1032454, "author_name": "kirderf", "author_url": "", "post_date": "09/30/2020 08:02:58", "content": "Thanks! Haha maybe can be the case, and will be even higher after this competition, many bronze medals and many many competitions just some points below the medalzone but still got ranking points, and also competitions with only ranking points as final prizes, well well, one have to be satisfied with that ranking after only one year with Kaggle :) Expert on being an Expert! ;)
", "votes": null, "replies": [] } ] }, { "id": 1032230, "author_name": "chankhavu", "author_url": "", "post_date": "09/30/2020 03:52:43", "content": "Congrats! May I ask you how you got such score just by modifying RANSAC parameters?
", "votes": null, "replies": [ { "id": 1032443, "author_name": "kirderf", "author_url": "", "post_date": "09/30/2020 07:53:49", "content": "Thanks, and same for you! Yes that was one part, RANSAC parameters, then also some other pre/post-processing, and used only the orginal pre-trained weight from the DELG-site, no extra training. Will make a write-up/summary later. Did though trained a couple of EffNets 4-7, but the limit of 40h/week on Kaggle TPU and 6h/day on Colab TPU, I should started the training much earlier in the competition to use them in final ensemble.
", "votes": null, "replies": [] } ] }, { "id": 1032436, "author_name": "andypenrose", "author_url": "", "post_date": "09/30/2020 07:48:00", "content": "Congratulations!
", "votes": null, "replies": [] }, { "id": 1033006, "author_name": "rahim3", "author_url": "", "post_date": "09/30/2020 15:37:56", "content": "Wow, great work 😍
\nCongratulations @kirderf on your first silver ✌️
Thanks! :) Yes hard work and lots of reading, it paid off. Now with 1y with Kaggle, with almost different challenges/problems every time I hope some will return my 2nd year, to face the task even better the second time with some experience :)
", "votes": null, "replies": [] }, { "id": 1033754, "author_name": "rahim3", "author_url": "", "post_date": "10/01/2020 08:40:20", "content": "Yes, if you keep trying and learning, one day it'll certainly pay off.
\nSomeday, we may be able to win a Gold ✌️
Congratulation! all I did so far is filling profile😅 your achievement seems really great😃
", "votes": null, "replies": [] }, { "id": 1035730, "author_name": "alincijov", "author_url": "", "post_date": "10/03/2020 02:50:04", "content": "Congratulations!
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1032168": "And that by using a singel DELG model+PP without ensemble, finally used the GLDv2 version to the baseline notebook, even using the GLDv1+PP resulted in a medal score. \n\nAnd congrats to the winners, what an achievement!\nIt will be valuable reading the write-ups.", "1032171": "Congrats!", "1032188": "Wow, Congrats ! U r probably the highest ranked Expert I have seen 😄", "1032230": "Congrats! May I ask you how you got such score just by modifying RANSAC parameters?", "1032436": "Congratulations!", "1032443": "Thanks, and same for you! Yes that was one part, RANSAC parameters, then also some other pre/post-processing, and used only the orginal pre-trained weight from the DELG-site, no extra training. Will make a write-up/summary later. Did though trained a couple of EffNets 4-7, but the limit of 40h/week on Kaggle TPU and 6h/day on Colab TPU, I should started the training much earlier in the competition to use them in final ensemble.", "1032454": "Thanks! Haha maybe can be the case, and will be even higher after this competition, many bronze medals and many many competitions just some points below the medalzone but still got ranking points, and also competitions with only ranking points as final prizes, well well, one have to be satisfied with that ranking after only one year with Kaggle :) Expert on being an Expert! ;)", "1032456": "Thanks, and same for you! :)", "1033006": "Wow, great work 😍\nCongratulations @kirderf on your first silver ✌️", "1033728": "Thanks! :) Yes hard work and lots of reading, it paid off. Now with 1y with Kaggle, with almost different challenges/problems every time I hope some will return my 2nd year, to face the task even better the second time with some experience :)", "1033754": "Yes, if you keep trying and learning, one day it'll certainly pay off. \nSomeday, we may be able to win a Gold ✌️", "1035730": "Congratulations!", "1040906": "Congratulation! all I did so far is filling profile😅 your achievement seems really great😃" }, "source": "meta" }