Hello, I decided to join this competition after reading the host's description:
\n\n\nIf successful, you'll help accelerate the world’s understanding of the relationships between cell and tissue organization. With a better idea of the relationship of cells, researchers will have more insight into the function of cells that impact human health. Further, the Human Reference Atlas constructed by HuBMAP will be freely available for use by researchers and pharmaceutical companies alike, potentially improving and prolonging human life.
\n
However it seems labels are sketchy to put it politely which then turns this competition from solving the problem of \" accelerating the world’s understanding of the relationships between cell and tissue organization\", into the problem of how to make a machine learning model perform good at a competition with poor labels; or mimic the (inconsistent) labeling process.
\nIf poor labeling has a reasonable cause (i.e. there are zillions of images, etc.) and this is advertised upfront in the data description (i.e. not all FTUs are labelled), I could empathize; however I think either Kaggle or the host needs to provide a clarification about the quality of the labels.
", "messages": [ { "id": 1899313, "postDate": "2022-08-15T07:28:49.983Z", "content": "Hello, I decided to join this competition after reading the host's description:
\n\n\nIf successful, you'll help accelerate the world’s understanding of the relationships between cell and tissue organization. With a better idea of the relationship of cells, researchers will have more insight into the function of cells that impact human health. Further, the Human Reference Atlas constructed by HuBMAP will be freely available for use by researchers and pharmaceutical companies alike, potentially improving and prolonging human life.
\n
However it seems labels are sketchy to put it politely which then turns this competition from solving the problem of \" accelerating the world’s understanding of the relationships between cell and tissue organization\", into the problem of how to make a machine learning model perform good at a competition with poor labels; or mimic the (inconsistent) labeling process.
\nIf poor labeling has a reasonable cause (i.e. there are zillions of images, etc.) and this is advertised upfront in the data description (i.e. not all FTUs are labelled), I could empathize; however I think either Kaggle or the host needs to provide a clarification about the quality of the labels.
", "rawMarkdown": "Hello, I decided to join this competition after reading the host's description:\n\n> If successful, you'll help accelerate the world’s understanding of the relationships between cell and tissue organization. With a better idea of the relationship of cells, researchers will have more insight into the function of cells that impact human health. Further, the Human Reference Atlas constructed by HuBMAP will be freely available for use by researchers and pharmaceutical companies alike, potentially improving and prolonging human life.\n\nHowever it seems[ labels are sketchy](https://www.kaggle.com/competitions/hubmap-organ-segmentation/discussion/344276) to put it politely which then turns this competition from solving the problem of \" accelerating the world’s understanding of the relationships between cell and tissue organization\", into the problem of how to make a machine learning model perform good at a competition with poor labels; or mimic the (inconsistent) labeling process.\n\nIf poor labeling has a reasonable cause (i.e. there are zillions of images, etc.) and this is advertised upfront in the data description (i.e. not all FTUs are labelled), I could empathize; however I think either **Kaggle or the host needs to provide a clarification about the quality of the labels**.", "votes": 28 }, { "id": 1922463, "postDate": "2022-09-01T13:56:38.273Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 1922463, "author_name": "", "author_url": "", "post_date": "2022-09-01T13:56:38.273000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1899313": "Hello, I decided to join this competition after reading the host's description:\n\n> If successful, you'll help accelerate the world’s understanding of the relationships between cell and tissue organization. With a better idea of the relationship of cells, researchers will have more insight into the function of cells that impact human health. Further, the Human Reference Atlas constructed by HuBMAP will be freely available for use by researchers and pharmaceutical companies alike, potentially improving and prolonging human life.\n\nHowever it seems[ labels are sketchy](https://www.kaggle.com/competitions/hubmap-organ-segmentation/discussion/344276) to put it politely which then turns this competition from solving the problem of \" accelerating the world’s understanding of the relationships between cell and tissue organization\", into the problem of how to make a machine learning model perform good at a competition with poor labels; or mimic the (inconsistent) labeling process.\n\nIf poor labeling has a reasonable cause (i.e. there are zillions of images, etc.) and this is advertised upfront in the data description (i.e. not all FTUs are labelled), I could empathize; however I think either **Kaggle or the host needs to provide a clarification about the quality of the labels**.", "1922463": "" } }