{ "id": 278801, "title": "šŸ¦ Broken masks?", "url": "/competitions/sartorius-cell-instance-segmentation/discussion/278801", "author_name": "Yaroslav Isaienkov", "post_date": "2021-10-15T17:46:45.504000", "votes": 37, "comment_count": 6, "views": 0, "content": "

Does anyone know what happened with some of the masks like in the example below (ex: blue in the top left corner)?
\nWhy do some have holes?

\n

Code to reproduce

\n

\"\"

", "messages": [ { "id": 1545929, "postDate": "2021-10-15T17:46:45.503Z", "content": "

Does anyone know what happened with some of the masks like in the example below (ex: blue in the top left corner)?
\nWhy do some have holes?

\n

Code to reproduce

\n

\"\"

", "rawMarkdown": "Does anyone know what happened with some of the masks like in the example below (ex: blue in the top left corner)? \nWhy do some have **holes**?\n\n[Code to reproduce](https://www.kaggle.com/ihelon/cell-segmentation-run-length-decoding)\n\n![](https://i.imgur.com/VV2RQiT.png)", "votes": 37 }, { "id": 1546813, "postDate": "2021-10-16T14:34:48.703Z", "content": "

I do believe its some kind of labeling error.
\nyou can easily fix this with scipy.ndimage.morphology.binary_fill_holes

", "rawMarkdown": "I do believe its some kind of labeling error. \nyou can easily fix this with scipy.ndimage.morphology.binary_fill_holes", "votes": 6, "replies": [ { "id": 1546981, "postDate": "2021-10-16T16:02:03.913Z", "content": "

The question is: if the test data has the same gaps which affects scoring.

", "rawMarkdown": "The question is: if the test data has the same gaps which affects scoring.", "votes": 5 }, { "id": 1547106, "postDate": "2021-10-16T17:09:58.797Z", "content": "

The annotation quality in itself is quite poor at this point.
\nMaybe we will get some new and corrected masks later in the challenge.
\nIf you take a look at the blue top left neuron, the additional area under the neuron is way worse than the small missing parts above. Maybe they had some conversion issues?

", "rawMarkdown": "The annotation quality in itself is quite poor at this point.\nMaybe we will get some new and corrected masks later in the challenge.\nIf you take a look at the blue top left neuron, the additional area under the neuron is way worse than the small missing parts above. Maybe they had some conversion issues?", "votes": 5 } ] }, { "id": 1548952, "postDate": "2021-10-18T16:51:38.110Z", "content": "

probable explanation : https://www.kaggle.com/c/sartorius-cell-instance-segmentation/discussion/279488

", "rawMarkdown": "probable explanation : https://www.kaggle.com/c/sartorius-cell-instance-segmentation/discussion/279488", "votes": 2 }, { "id": 1547226, "postDate": "2021-10-16T22:42:47.647Z", "content": "

And we have so similar pattern for missing parts inside objects. It misses only in the top border part of the figures.

", "rawMarkdown": "And we have so similar pattern for missing parts inside objects. It misses only in the top border part of the figures." }, { "id": 1550683, "postDate": "2021-10-19T23:26:20.697Z", "rawMarkdown": "", "votes": -2, "isDeleted": true } ], "comments": [ { "id": 1546813, "author_name": "Philipp Sodmann", "author_url": "", "post_date": "2021-10-16T14:34:48.703000", "content": "

I do believe its some kind of labeling error.
\nyou can easily fix this with scipy.ndimage.morphology.binary_fill_holes

", "votes": 6, "replies": [ { "id": 1546981, "author_name": "Valentin Nikotin", "author_url": "", "post_date": "2021-10-16T16:02:03.913000", "content": "

The question is: if the test data has the same gaps which affects scoring.

", "votes": 5, "replies": [] }, { "id": 1547106, "author_name": "Philipp Sodmann", "author_url": "", "post_date": "2021-10-16T17:09:58.797000", "content": "

The annotation quality in itself is quite poor at this point.
\nMaybe we will get some new and corrected masks later in the challenge.
\nIf you take a look at the blue top left neuron, the additional area under the neuron is way worse than the small missing parts above. Maybe they had some conversion issues?

", "votes": 5, "replies": [] } ] }, { "id": 1548952, "author_name": "Gleb", "author_url": "", "post_date": "2021-10-18T16:51:38.110000", "content": "

probable explanation : https://www.kaggle.com/c/sartorius-cell-instance-segmentation/discussion/279488

", "votes": 2, "replies": [] }, { "id": 1547226, "author_name": "Yaroslav Isaienkov", "author_url": "", "post_date": "2021-10-16T22:42:47.647000", "content": "

And we have so similar pattern for missing parts inside objects. It misses only in the top border part of the figures.

", "votes": 0, "replies": [] }, { "id": 1550683, "author_name": "", "author_url": "", "post_date": "2021-10-19T23:26:20.697000", "content": "", "votes": -2, "replies": [] } ], "raw_markdown_by_id": { "1545929": "Does anyone know what happened with some of the masks like in the example below (ex: blue in the top left corner)? \nWhy do some have **holes**?\n\n[Code to reproduce](https://www.kaggle.com/ihelon/cell-segmentation-run-length-decoding)\n\n![](https://i.imgur.com/VV2RQiT.png)", "1546813": "I do believe its some kind of labeling error. \nyou can easily fix this with scipy.ndimage.morphology.binary_fill_holes", "1548952": "probable explanation : https://www.kaggle.com/c/sartorius-cell-instance-segmentation/discussion/279488", "1547226": "And we have so similar pattern for missing parts inside objects. It misses only in the top border part of the figures.", "1550683": "" } }