{ "id": 531771, "title": "Human error ???", "url": "/competitions/rsna-2024-lumbar-spine-degenerative-classification/discussion/531771", "author_name": "", "post_date": "2024-09-03T02:47:41.582576Z", "votes": 13, "comment_count": 6, "views": 0, "content": "

Here is an image showing the error ground truth given.
\nI don't think human make such a mistake.

\n

Hence i think the labeling process could be:

\n\n

\"\"

", "messages": [ { "id": "2977602", "postDate": "09/03/2024 02:47:41", "content": "

Here is an image showing the error ground truth given.
\nI don't think human make such a mistake.

\n

Hence i think the labeling process could be:

\n\n

\"\"

", "rawMarkdown": "Here is an image showing the error ground truth given.\nI don't think human make such a mistake.\n\nHence i think the labeling process could be:\n- lablled by some machine model\n- verified by human annotator\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F104e641bfd19673a706f6e1690a54972%2FSelection_999(5992).png?generation=1725331578524493&alt=media)", "votes": null }, { "id": "2977735", "postDate": "09/03/2024 06:53:01", "content": "

I also agree with you. Aside from that, I think there are many wrong coordinates in foraminal narrowing.
\n In some samples (ex study id: 3480260143), there is a huge difference in left and right coordinates at the same disk level.

", "rawMarkdown": "I also agree with you. Aside from that, I think there are many wrong coordinates in foraminal narrowing.\n In some samples (ex study id: 3480260143), there is a huge difference in left and right coordinates at the same disk level.", "votes": null }, { "id": "2979563", "postDate": "09/04/2024 22:25:56", "content": "

It has to be human annotated, looking at how accurate it was for the foramina

", "rawMarkdown": "It has to be human annotated, looking at how accurate it was for the foramina", "votes": null }, { "id": "2981011", "postDate": "09/06/2024 12:03:20", "content": "

If the labelling was done on axial, then I could see how the mistake could be made. Not on sagittal though

", "rawMarkdown": "If the labelling was done on axial, then I could see how the mistake could be made. Not on sagittal though", "votes": null }, { "id": "2992906", "postDate": "09/19/2024 08:28:09", "content": "

Hope these kind of mistakes are not there in hidden test set otherwise good model will get bad scores.

", "rawMarkdown": "Hope these kind of mistakes are not there in hidden test set otherwise good model will get bad scores.", "votes": null }, { "id": "2992975", "postDate": "09/19/2024 10:13:59", "content": "

you can do a probe to see if hidden contains these.
\nyou need to study the kaggle train data and see how you can detect misalignment label from your prediction

", "rawMarkdown": "you can do a probe to see if hidden contains these.\nyou need to study the kaggle train data and see how you can detect misalignment label from your prediction", "votes": null }, { "id": "2997434", "postDate": "09/24/2024 14:33:11", "content": "

Is this error random or consistent?

", "rawMarkdown": "Is this error random or consistent?", "votes": null } ], "comments": [ { "id": 2977735, "author_name": "bmejspark", "author_url": "", "post_date": "09/03/2024 06:53:01", "content": "

I also agree with you. Aside from that, I think there are many wrong coordinates in foraminal narrowing.
\n In some samples (ex study id: 3480260143), there is a huge difference in left and right coordinates at the same disk level.

", "votes": null, "replies": [ { "id": 2979563, "author_name": "theexaltedone", "author_url": "", "post_date": "09/04/2024 22:25:56", "content": "

It has to be human annotated, looking at how accurate it was for the foramina

", "votes": null, "replies": [] } ] }, { "id": 2981011, "author_name": "reubenschmidt", "author_url": "", "post_date": "09/06/2024 12:03:20", "content": "

If the labelling was done on axial, then I could see how the mistake could be made. Not on sagittal though

", "votes": null, "replies": [] }, { "id": 2992906, "author_name": "rohitchaudhari25", "author_url": "", "post_date": "09/19/2024 08:28:09", "content": "

Hope these kind of mistakes are not there in hidden test set otherwise good model will get bad scores.

", "votes": null, "replies": [ { "id": 2992975, "author_name": "hengck23", "author_url": "", "post_date": "09/19/2024 10:13:59", "content": "

you can do a probe to see if hidden contains these.
\nyou need to study the kaggle train data and see how you can detect misalignment label from your prediction

", "votes": null, "replies": [] } ] }, { "id": 2997434, "author_name": "", "author_url": "", "post_date": "09/24/2024 14:33:11", "content": "

Is this error random or consistent?

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2977602": "Here is an image showing the error ground truth given.\nI don't think human make such a mistake.\n\nHence i think the labeling process could be:\n- lablled by some machine model\n- verified by human annotator\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F104e641bfd19673a706f6e1690a54972%2FSelection_999(5992).png?generation=1725331578524493&alt=media)", "2977735": "I also agree with you. Aside from that, I think there are many wrong coordinates in foraminal narrowing.\n In some samples (ex study id: 3480260143), there is a huge difference in left and right coordinates at the same disk level.", "2979563": "It has to be human annotated, looking at how accurate it was for the foramina", "2981011": "If the labelling was done on axial, then I could see how the mistake could be made. Not on sagittal though", "2992906": "Hope these kind of mistakes are not there in hidden test set otherwise good model will get bad scores.", "2992975": "you can do a probe to see if hidden contains these.\nyou need to study the kaggle train data and see how you can detect misalignment label from your prediction", "2997434": "Is this error random or consistent?" }, "source": "meta" }