{ "id": 230802, "title": "Using Cellpose for cell segmentation", "url": "/competitions/hpa-single-cell-image-classification/discussion/230802", "author_name": "Andrew Scholan", "post_date": "2021-04-05T18:26:01.931000", "votes": 10, "comment_count": 0, "views": 0, "content": "

Cellpose is one of the libraries for cell segmentation suggested by Emma Lundberg in her introductory discussion Human Protein Atlas - Single Cell Classification.

\n

Unfortunately cellpose is not in the standard notebook configuration and requires internet access to load in the pre-trained model weights. This means that it's hard to use in the HPA21 submission notebooks.

\n

However, if anyone is interested in using this library I have shared the following notebooks that get round the technical problems. Initial EDA shows that cellpose works pretty well for segmentation (beware… very small sample size!!!!).

\n
    \n
  1. Notebook that shows how to package up the cellpose library so that it can be offline loaded as a Kaggle dataset: Offline Package Wheeler
    2. \n
  2. Notebook that shows how to package up the cellpose pre-trained models into a Kaggle dataset: Cellpose Model Collector
    3. \n
  3. Notebook that pulls it all together and shows how to use cellpose for some basic segmentation: HPA21 Cellpose EDA
    4. \n
\n

Cheers,
\nAndrew.

", "messages": [ { "id": 1263869, "postDate": "2021-04-05T18:26:01.930Z", "content": "

Cellpose is one of the libraries for cell segmentation suggested by Emma Lundberg in her introductory discussion Human Protein Atlas - Single Cell Classification.

\n

Unfortunately cellpose is not in the standard notebook configuration and requires internet access to load in the pre-trained model weights. This means that it's hard to use in the HPA21 submission notebooks.

\n

However, if anyone is interested in using this library I have shared the following notebooks that get round the technical problems. Initial EDA shows that cellpose works pretty well for segmentation (beware… very small sample size!!!!).

\n
    \n
  1. Notebook that shows how to package up the cellpose library so that it can be offline loaded as a Kaggle dataset: Offline Package Wheeler
    2. \n
  2. Notebook that shows how to package up the cellpose pre-trained models into a Kaggle dataset: Cellpose Model Collector
    3. \n
  3. Notebook that pulls it all together and shows how to use cellpose for some basic segmentation: HPA21 Cellpose EDA
    4. \n
\n

Cheers,
\nAndrew.

", "rawMarkdown": "[Cellpose ](https://www.cellpose.org/) is one of the libraries for cell segmentation suggested by [Emma Lundberg](https://www.kaggle.com/emmalumpan) in her introductory discussion [Human Protein Atlas - Single Cell Classification](https://www.kaggle.com/c/hpa-single-cell-image-classification/discussion/214616).\n\nUnfortunately cellpose is not in the standard notebook configuration and requires internet access to load in the pre-trained model weights. This means that it's hard to use in the HPA21 submission notebooks.\n\nHowever, if anyone is interested in using this library I have shared the following notebooks that get round the technical problems. Initial EDA shows that cellpose works pretty well for segmentation (beware... very small sample size!!!!).\n\n1. Notebook that shows how to package up the cellpose library so that it can be offline loaded as a Kaggle dataset: [Offline Package Wheeler](https://www.kaggle.com/andrewscholan/offline-package-wheeler-public)\n2. Notebook that shows how to package up the cellpose pre-trained models into a Kaggle dataset: [Cellpose Model Collector](https://www.kaggle.com/andrewscholan/cellpose-model-collector-public)\n3. Notebook that pulls it all together and shows how to use cellpose for some basic segmentation: [HPA21 Cellpose EDA](https://www.kaggle.com/andrewscholan/hpa21-cellpose-eda-public)\n\nCheers,\nAndrew.\n\n", "votes": 10 } ], "comments": [], "raw_markdown_by_id": { "1263869": "[Cellpose ](https://www.cellpose.org/) is one of the libraries for cell segmentation suggested by [Emma Lundberg](https://www.kaggle.com/emmalumpan) in her introductory discussion [Human Protein Atlas - Single Cell Classification](https://www.kaggle.com/c/hpa-single-cell-image-classification/discussion/214616).\n\nUnfortunately cellpose is not in the standard notebook configuration and requires internet access to load in the pre-trained model weights. This means that it's hard to use in the HPA21 submission notebooks.\n\nHowever, if anyone is interested in using this library I have shared the following notebooks that get round the technical problems. Initial EDA shows that cellpose works pretty well for segmentation (beware... very small sample size!!!!).\n\n1. Notebook that shows how to package up the cellpose library so that it can be offline loaded as a Kaggle dataset: [Offline Package Wheeler](https://www.kaggle.com/andrewscholan/offline-package-wheeler-public)\n2. Notebook that shows how to package up the cellpose pre-trained models into a Kaggle dataset: [Cellpose Model Collector](https://www.kaggle.com/andrewscholan/cellpose-model-collector-public)\n3. Notebook that pulls it all together and shows how to use cellpose for some basic segmentation: [HPA21 Cellpose EDA](https://www.kaggle.com/andrewscholan/hpa21-cellpose-eda-public)\n\nCheers,\nAndrew.\n\n" } }