{ "id": 200141, "title": "[completed] starter kit ....based on 2019 kaggle cassava challenge", "url": "/competitions/cassava-leaf-disease-classification/discussion/200141", "author_name": "", "post_date": "2020-11-29T03:43:28.575597900Z", "votes": 29, "comment_count": 6, "views": 0, "content": "

done!

\n

I spend some time to repeat the top solutions from the 2019 kaggle cassava challenge
\nhttps://www.kaggle.com/c/cassava-disease/overview

\n

This is for me to check my code and understanding of the 2019 solutions. It also enables me to understand the similarities and differences of the two datasets from 2019 and 2020 (this current challenge)

\n

here are the results

\n

\"\"

\n

\"\"

\n

code and models can be downloaded at:
\nhttps://drive.google.com/drive/folders/1IwQ9gPA2ltVgf0DrWUb-Jk47FN0M7vyh?usp=sharing

\n

the trained model is compatible with my submission kernel:
\nhttps://www.kaggle.com/hengck23/notebook6262cd4cd3

", "messages": [ { "id": "1094875", "postDate": "11/29/2020 03:43:28", "content": "

done!

\n

I spend some time to repeat the top solutions from the 2019 kaggle cassava challenge
\nhttps://www.kaggle.com/c/cassava-disease/overview

\n

This is for me to check my code and understanding of the 2019 solutions. It also enables me to understand the similarities and differences of the two datasets from 2019 and 2020 (this current challenge)

\n

here are the results

\n

\"\"

\n

\"\"

\n

code and models can be downloaded at:
\nhttps://drive.google.com/drive/folders/1IwQ9gPA2ltVgf0DrWUb-Jk47FN0M7vyh?usp=sharing

\n

the trained model is compatible with my submission kernel:
\nhttps://www.kaggle.com/hengck23/notebook6262cd4cd3

", "rawMarkdown": "done!\n\nI spend some time to repeat the top solutions from the 2019 kaggle cassava challenge\nhttps://www.kaggle.com/c/cassava-disease/overview\n\nThis is for me to check my code and understanding of the 2019 solutions. It also enables me to understand the similarities and differences of the two datasets from 2019 and 2020 (this current challenge)\n\nhere are the results\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F354f0f05de433854c53f88acc03ebbf3%2FSelection_080.png?generation=1606621327031700&alt=media)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F89c60081120b8ad016bdd0765cd58849%2FSelection_079.png?generation=1606621353915133&alt=media)\n\n\n\ncode and models can be downloaded at:\nhttps://drive.google.com/drive/folders/1IwQ9gPA2ltVgf0DrWUb-Jk47FN0M7vyh?usp=sharing\n\n\nthe trained model is compatible with my submission kernel:\nhttps://www.kaggle.com/hengck23/notebook6262cd4cd3", "votes": null }, { "id": "1095986", "postDate": "11/30/2020 05:56:16", "content": "

Nice work, in 1st place discussion when they say using image above 95% probability are they referring to psuedolabeling extra images and using it for training then checkin CV?

", "rawMarkdown": "Nice work, in 1st place discussion when they say using image above 95% probability are they referring to psuedolabeling extra images and using it for training then checkin CV?", "votes": null }, { "id": "1104251", "postDate": "12/06/2020 18:46:58", "content": "

Thank you for sharing your experiments @hengck23! Have you tried also to use the unlabelled images (12k samples) from 2019 comp. ? either for train or extra valid set

", "rawMarkdown": "Thank you for sharing your experiments @hengck23! Have you tried also to use the unlabelled images (12k samples) from 2019 comp. ? either for train or extra valid set", "votes": null }, { "id": "1104439", "postDate": "12/07/2020 00:23:31", "content": "

Thanks for sharing your investigation.

", "rawMarkdown": "Thanks for sharing your investigation.", "votes": null }, { "id": "1107389", "postDate": "12/09/2020 16:51:35", "content": "

Nice work! Thanks for sharing these posts.

", "rawMarkdown": "Nice work! Thanks for sharing these posts.", "votes": null }, { "id": "1109308", "postDate": "12/11/2020 14:20:00", "content": "

Nice! @hengck23 have you tried to use the same training pipeline on this competition to see what results you would get?

", "rawMarkdown": "Nice! @hengck23 have you tried to use the same training pipeline on this competition to see what results you would get?", "votes": null }, { "id": "1109344", "postDate": "12/11/2020 15:03:30", "content": "

my current results is based on this training kit. the classifier, loss etc is changed

", "rawMarkdown": "my current results is based on this training kit. the classifier, loss etc is changed", "votes": null } ], "comments": [ { "id": 1095986, "author_name": "keremt", "author_url": "", "post_date": "11/30/2020 05:56:16", "content": "

Nice work, in 1st place discussion when they say using image above 95% probability are they referring to psuedolabeling extra images and using it for training then checkin CV?

", "votes": null, "replies": [] }, { "id": 1104251, "author_name": "imeintanis", "author_url": "", "post_date": "12/06/2020 18:46:58", "content": "

Thank you for sharing your experiments @hengck23! Have you tried also to use the unlabelled images (12k samples) from 2019 comp. ? either for train or extra valid set

", "votes": null, "replies": [] }, { "id": 1104439, "author_name": "luqing2", "author_url": "", "post_date": "12/07/2020 00:23:31", "content": "

Thanks for sharing your investigation.

", "votes": null, "replies": [] }, { "id": 1107389, "author_name": "aikeyz", "author_url": "", "post_date": "12/09/2020 16:51:35", "content": "

Nice work! Thanks for sharing these posts.

", "votes": null, "replies": [] }, { "id": 1109308, "author_name": "dimitreoliveira", "author_url": "", "post_date": "12/11/2020 14:20:00", "content": "

Nice! @hengck23 have you tried to use the same training pipeline on this competition to see what results you would get?

", "votes": null, "replies": [ { "id": 1109344, "author_name": "hengck23", "author_url": "", "post_date": "12/11/2020 15:03:30", "content": "

my current results is based on this training kit. the classifier, loss etc is changed

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1094875": "done!\n\nI spend some time to repeat the top solutions from the 2019 kaggle cassava challenge\nhttps://www.kaggle.com/c/cassava-disease/overview\n\nThis is for me to check my code and understanding of the 2019 solutions. It also enables me to understand the similarities and differences of the two datasets from 2019 and 2020 (this current challenge)\n\nhere are the results\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F354f0f05de433854c53f88acc03ebbf3%2FSelection_080.png?generation=1606621327031700&alt=media)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F89c60081120b8ad016bdd0765cd58849%2FSelection_079.png?generation=1606621353915133&alt=media)\n\n\n\ncode and models can be downloaded at:\nhttps://drive.google.com/drive/folders/1IwQ9gPA2ltVgf0DrWUb-Jk47FN0M7vyh?usp=sharing\n\n\nthe trained model is compatible with my submission kernel:\nhttps://www.kaggle.com/hengck23/notebook6262cd4cd3", "1095986": "Nice work, in 1st place discussion when they say using image above 95% probability are they referring to psuedolabeling extra images and using it for training then checkin CV?", "1104251": "Thank you for sharing your experiments @hengck23! Have you tried also to use the unlabelled images (12k samples) from 2019 comp. ? either for train or extra valid set", "1104439": "Thanks for sharing your investigation.", "1107389": "Nice work! Thanks for sharing these posts.", "1109308": "Nice! @hengck23 have you tried to use the same training pipeline on this competition to see what results you would get?", "1109344": "my current results is based on this training kit. the classifier, loss etc is changed" }, "source": "meta" }