![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F113660%2F4e8d13412e9b9834fcdfdbde439829ce%2FSlide5.png?generation=1572936504805578&alt=media\")
now we have ground truth of stage_1 test, let's us see how well they perform. This is for stage1 submission of LB =0.067
\n\n
I wonder is there any label noise below (the label are not continuous)?
\n\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F113660%2Fd48011362f64536a1f7fd89f401061f6%2FSlide3.png?generation=1572936501271810&alt=media\")
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F113660%2F8d48d578ca5a006b04f782f4b056f145%2FSlide1.png?generation=1572936503502035&alt=media\")
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F113660%2F9349a57b78138aa5691d4321d2bf3f3f%2FSlide4.png?generation=1572936506626350&alt=media\")
now we have ground truth of stage_1 test, let's us see how well they perform. This is for stage1 submission of LB =0.067
\n\n
I wonder is there any label noise below (the label are not continuous)?
\n\n\n\n
here is the PPT for more comparsion
", "rawMarkdown": "here is the PPT for more comparsion", "votes": 3 }, { "id": 665617, "postDate": "2019-11-05T07:23:08.917Z", "content": "apparently, they are not label noise. \n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F113660%2F99313660d48b3b50767d3d2c085719c1%2FSelection_059.png?generation=1572938586385599&alt=media\")
my groupby method is wrong. i use:
\n\n```\ngb = df.groupby(['patient_id','study_instance_id'])
\n\nthe example above is for\npatient_id = ID_fe2022a4
\nstudy_instance_id = ID_aec2f7688c
```
", "rawMarkdown": "apparently, they are not label noise. \n\n\n\n\nmy groupby method is wrong. i use:\n\n```\ngb = df.groupby(['patient_id','study_instance_id'])\n\nthe example above is for\npatient_id = ID_fe2022a4\t\nstudy_instance_id = ID_aec2f7688c\n\n\n```", "votes": 1, "replies": [ { "id": 666376, "postDate": "2019-11-06T03:40:40.093Z", "content": "such overlap instance can be detected by checking the inconsistency of the slice depth interval in the group.
\n\na cleanup after groupby should be fine
", "rawMarkdown": "such overlap instance can be detected by checking the inconsistency of the slice depth interval in the group.\n\na cleanup after groupby should be fine" } ] }, { "id": 667808, "postDate": "2019-11-07T17:52:17.390Z", "content": "ID_8d740f970, ID_adc8b8c5f and ID_4156734aa, according to the images shown, definitely have intraventricular hemorrhage and most probably SAH.
", "rawMarkdown": "ID_8d740f970, ID_adc8b8c5f and ID_4156734aa, according to the images shown, definitely have intraventricular hemorrhage and most probably SAH." }, { "id": 665675, "postDate": "2019-11-05T09:21:38.570Z", "content": "Thanks, you are amazing.
", "rawMarkdown": "Thanks, you are amazing." } ], "comments": [ { "id": 665598, "author_name": "hengck23", "author_url": "", "post_date": "2019-11-05T06:54:41.383000", "content": "here is the PPT for more comparsion
", "votes": 3, "replies": [] }, { "id": 665617, "author_name": "hengck23", "author_url": "", "post_date": "2019-11-05T07:23:08.917000", "content": "apparently, they are not label noise. \n
my groupby method is wrong. i use:
\n\n```\ngb = df.groupby(['patient_id','study_instance_id'])
\n\nthe example above is for\npatient_id = ID_fe2022a4
\nstudy_instance_id = ID_aec2f7688c
```
", "votes": 1, "replies": [ { "id": 666376, "author_name": "hengck23", "author_url": "", "post_date": "2019-11-06T03:40:40.093000", "content": "such overlap instance can be detected by checking the inconsistency of the slice depth interval in the group.
\n\na cleanup after groupby should be fine
", "votes": 0, "replies": [] } ] }, { "id": 667808, "author_name": "Alexandre Cadrin-Chênevert", "author_url": "", "post_date": "2019-11-07T17:52:17.390000", "content": "ID_8d740f970, ID_adc8b8c5f and ID_4156734aa, according to the images shown, definitely have intraventricular hemorrhage and most probably SAH.
", "votes": 0, "replies": [] }, { "id": 665675, "author_name": "LongYin/杰少", "author_url": "", "post_date": "2019-11-05T09:21:38.570000", "content": "Thanks, you are amazing.
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "665596": "now we have ground truth of stage\\_1 test, let's us see how well they perform. This is for stage1 submission of LB =0.067\n\n\n\n\nI wonder is there any label noise below (the label are not continuous)?\n\n\n\n\n\n", "665598": "here is the PPT for more comparsion", "665617": "apparently, they are not label noise. \n\n\n\n\nmy groupby method is wrong. i use:\n\n```\ngb = df.groupby(['patient_id','study_instance_id'])\n\nthe example above is for\npatient_id = ID_fe2022a4\t\nstudy_instance_id = ID_aec2f7688c\n\n\n```", "667808": "ID_8d740f970, ID_adc8b8c5f and ID_4156734aa, according to the images shown, definitely have intraventricular hemorrhage and most probably SAH.", "665675": "Thanks, you are amazing." } }