{ "id": 224539, "title": "training with additional datasets", "url": "/competitions/vinbigdata-chest-xray-abnormalities-detection/discussion/224539", "author_name": "Wonjun Park", "post_date": "2021-03-09T01:05:24.678000", "votes": 4, "comment_count": 4, "views": 0, "content": "

In a previous post, I have organized additional datasets.

\n\n

Among them, the only dataset available for detection tasks was NIH dataset labeled with bbox.

\n

It showed some performance improvements when trained based on the detectron2 model.

\n

If you want to see more information about adding data, we recommend checking version 8 of the kernel below.

\n", "messages": [ { "id": 1231416, "postDate": "2021-03-09T01:05:24.680Z", "content": "

In a previous post, I have organized additional datasets.

\n\n

Among them, the only dataset available for detection tasks was NIH dataset labeled with bbox.

\n

It showed some performance improvements when trained based on the detectron2 model.

\n

If you want to see more information about adding data, we recommend checking version 8 of the kernel below.

\n", "rawMarkdown": "In a previous post, I have organized additional datasets.\n- https://www.kaggle.com/c/vinbigdata-chest-xray-abnormalities-detection/discussion/215792\n- https://www.kaggle.com/c/vinbigdata-chest-xray-abnormalities-detection/discussion/215793\n\nAmong them, the only dataset available for detection tasks was NIH dataset labeled with bbox.\n\nIt showed some performance improvements when trained based on the detectron2 model.\n\nIf you want to see more information about adding data, we recommend checking version 8 of the kernel below.\n\n- https://www.kaggle.com/wonjunpark/vinbigdata-detectron2-train-using-nih-dataset", "votes": 4 }, { "id": 1234277, "postDate": "2021-03-11T05:19:31.067Z", "content": "

@wojunpark

\n

Thanks for the extra dataset pointers

\n

May I ask how much improvement do you observe for pretraining with NIH data ?

", "rawMarkdown": "@wojunpark \n\nThanks for the extra dataset pointers\n\nMay I ask how much improvement do you observe for pretraining with NIH data ? ", "votes": 1, "replies": [ { "id": 1235316, "postDate": "2021-03-12T03:42:15.490Z", "content": "

x.x84 -> x.x85

", "rawMarkdown": "x.x84 -> x.x85" } ] }, { "id": 1231965, "postDate": "2021-03-09T12:00:02.467Z", "content": "

@wojunpark Thanks for sharing the additional datasets used and the notebook ..Super Helpful

", "rawMarkdown": "@wojunpark Thanks for sharing the additional datasets used and the notebook ..Super Helpful", "votes": 1, "replies": [ { "id": 1235315, "postDate": "2021-03-12T03:40:46.340Z", "content": "

Thank you. I'm glad you were helpful.

", "rawMarkdown": "Thank you. I'm glad you were helpful." } ] } ], "comments": [ { "id": 1234277, "author_name": "NakedKoala", "author_url": "", "post_date": "2021-03-11T05:19:31.067000", "content": "

@wojunpark

\n

Thanks for the extra dataset pointers

\n

May I ask how much improvement do you observe for pretraining with NIH data ?

", "votes": 1, "replies": [ { "id": 1235316, "author_name": "Wonjun Park", "author_url": "", "post_date": "2021-03-12T03:42:15.490000", "content": "

x.x84 -> x.x85

", "votes": 0, "replies": [] } ] }, { "id": 1231965, "author_name": "Tensor Girl", "author_url": "", "post_date": "2021-03-09T12:00:02.467000", "content": "

@wojunpark Thanks for sharing the additional datasets used and the notebook ..Super Helpful

", "votes": 1, "replies": [ { "id": 1235315, "author_name": "Wonjun Park", "author_url": "", "post_date": "2021-03-12T03:40:46.340000", "content": "

Thank you. I'm glad you were helpful.

", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "1231416": "In a previous post, I have organized additional datasets.\n- https://www.kaggle.com/c/vinbigdata-chest-xray-abnormalities-detection/discussion/215792\n- https://www.kaggle.com/c/vinbigdata-chest-xray-abnormalities-detection/discussion/215793\n\nAmong them, the only dataset available for detection tasks was NIH dataset labeled with bbox.\n\nIt showed some performance improvements when trained based on the detectron2 model.\n\nIf you want to see more information about adding data, we recommend checking version 8 of the kernel below.\n\n- https://www.kaggle.com/wonjunpark/vinbigdata-detectron2-train-using-nih-dataset", "1234277": "@wojunpark \n\nThanks for the extra dataset pointers\n\nMay I ask how much improvement do you observe for pretraining with NIH data ? ", "1231965": "@wojunpark Thanks for sharing the additional datasets used and the notebook ..Super Helpful" } }