{ "id": 348241, "title": "Vertebrae detection with 95% accuracy", "url": "/competitions/rsna-2022-cervical-spine-fracture-detection/discussion/348241", "author_name": "", "post_date": "2022-08-27T16:29:08.072130800Z", "votes": 36, "comment_count": 5, "views": 0, "content": "

Vertebrae detection is an essential block in the final solution of this challenge. Many of you have seen great work of Samuel Cortinhas Extracting Vertebrae C1, …, C7 where vertebraes are detected using a simple RandomForestClassifier with 88% accuracy.

\n

Here I want to share an improved vertebrae detector and a dataset with vertebrae detections of all train images PyTorch-EffNetV2 vertebrae detection (acc: 0.94). Predicted C1-C7 labels can be used to help training the main fracture detection model. Trained vertebrae detection models can also be used directly in your submissions.

\n

Hope you find it useful.

", "messages": [ { "id": "1916107", "postDate": "08/27/2022 16:29:08", "content": "

Vertebrae detection is an essential block in the final solution of this challenge. Many of you have seen great work of Samuel Cortinhas Extracting Vertebrae C1, …, C7 where vertebraes are detected using a simple RandomForestClassifier with 88% accuracy.

\n

Here I want to share an improved vertebrae detector and a dataset with vertebrae detections of all train images PyTorch-EffNetV2 vertebrae detection (acc: 0.94). Predicted C1-C7 labels can be used to help training the main fracture detection model. Trained vertebrae detection models can also be used directly in your submissions.

\n

Hope you find it useful.

", "rawMarkdown": "Vertebrae detection is an essential block in the final solution of this challenge. Many of you have seen great work of Samuel Cortinhas [Extracting Vertebrae C1, ..., C7](https://www.kaggle.com/code/samuelcortinhas/extracting-vertebrae-c1-c7) where vertebraes are detected using a simple RandomForestClassifier with 88% accuracy. \n\nHere I want to share an improved vertebrae detector and a dataset with vertebrae detections of all train images [PyTorch-EffNetV2 vertebrae detection (acc: 0.94)](https://www.kaggle.com/vslaykovsky/pytorch-effnetv2-vertebrae-detection-acc-0-95). Predicted C1-C7 labels can be used to help training the main fracture detection model. Trained vertebrae detection models can also be used directly in your submissions. \n\nHope you find it useful.", "votes": null }, { "id": "1916294", "postDate": "08/27/2022 18:25:15", "content": "

Upvoted and bookmarked for later study. Thanks, @vslaykovsky, much appreciated!

", "rawMarkdown": "Upvoted and bookmarked for later study. Thanks, @vslaykovsky, much appreciated!", "votes": null }, { "id": "1918507", "postDate": "08/29/2022 16:13:48", "content": "

Quite clever. I never would have guessed that you can predict the type of vertebrae that accurately from images.

", "rawMarkdown": "Quite clever. I never would have guessed that you can predict the type of vertebrae that accurately from images.", "votes": null }, { "id": "1923699", "postDate": "09/02/2022 13:17:38", "content": "

I don't know why you join train_detection.csv and org_train.csv could you explain for me?

", "rawMarkdown": "I don't know why you join train_detection.csv and org_train.csv could you explain for me?", "votes": null }, { "id": "1925792", "postDate": "09/04/2022 10:07:03", "content": "

Useful topic bookmarked for later. Thanks

", "rawMarkdown": "Useful topic bookmarked for later. Thanks", "votes": null }, { "id": "1958000", "postDate": "09/27/2022 08:43:45", "content": "

Great Work

", "rawMarkdown": "Great Work", "votes": null } ], "comments": [ { "id": 1916294, "author_name": "anityagangurde", "author_url": "", "post_date": "08/27/2022 18:25:15", "content": "

Upvoted and bookmarked for later study. Thanks, @vslaykovsky, much appreciated!

", "votes": null, "replies": [] }, { "id": 1918507, "author_name": "solverworld", "author_url": "", "post_date": "08/29/2022 16:13:48", "content": "

Quite clever. I never would have guessed that you can predict the type of vertebrae that accurately from images.

", "votes": null, "replies": [] }, { "id": 1923699, "author_name": "kalelpark", "author_url": "", "post_date": "09/02/2022 13:17:38", "content": "

I don't know why you join train_detection.csv and org_train.csv could you explain for me?

", "votes": null, "replies": [] }, { "id": 1925792, "author_name": "erolmasimov", "author_url": "", "post_date": "09/04/2022 10:07:03", "content": "

Useful topic bookmarked for later. Thanks

", "votes": null, "replies": [] }, { "id": 1958000, "author_name": "darshan3k", "author_url": "", "post_date": "09/27/2022 08:43:45", "content": "

Great Work

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1916107": "Vertebrae detection is an essential block in the final solution of this challenge. Many of you have seen great work of Samuel Cortinhas [Extracting Vertebrae C1, ..., C7](https://www.kaggle.com/code/samuelcortinhas/extracting-vertebrae-c1-c7) where vertebraes are detected using a simple RandomForestClassifier with 88% accuracy. \n\nHere I want to share an improved vertebrae detector and a dataset with vertebrae detections of all train images [PyTorch-EffNetV2 vertebrae detection (acc: 0.94)](https://www.kaggle.com/vslaykovsky/pytorch-effnetv2-vertebrae-detection-acc-0-95). Predicted C1-C7 labels can be used to help training the main fracture detection model. Trained vertebrae detection models can also be used directly in your submissions. \n\nHope you find it useful.", "1916294": "Upvoted and bookmarked for later study. Thanks, @vslaykovsky, much appreciated!", "1918507": "Quite clever. I never would have guessed that you can predict the type of vertebrae that accurately from images.", "1923699": "I don't know why you join train_detection.csv and org_train.csv could you explain for me?", "1925792": "Useful topic bookmarked for later. Thanks", "1958000": "Great Work" }, "source": "meta" }