{ "id": 215428, "title": "Updates coming!", "url": "/competitions/hpa-single-cell-image-classification/discussion/215428", "author_name": "Phil Culliton", "post_date": "2021-01-29T18:51:51.404000", "votes": 34, "comment_count": 9, "views": 0, "content": "

Hi all -

\n

We've had some excellent early feedback in this competition - thank you to everyone participating so far! This competition poses an interesting problem, and we will be taking some steps to improve the experience of competing.

\n

Over the next few business days we will be:

\n
    \n
  1. Shifting from F1 to mAP, which should be easier to use / more readily available. (The new metric should be precisely the OpenImages mAP metric mentioned in the Evaluation tab.)
    2. \n
  2. Updating the 8-bit PNGs to 16-bit, which should be useful for modeling.
    3. \n
  3. Updating the public dataset to make it easier / faster to download.
    4. \n
\n

Note that when #1 and #2 are complete we will be re-scoring submissions.

\n

Thanks again for all your feedback, and please let us know if you have questions. We will post messages about updates as they occur.

", "messages": [ { "id": 1176726, "postDate": "2021-01-29T18:51:51.403Z", "content": "

Hi all -

\n

We've had some excellent early feedback in this competition - thank you to everyone participating so far! This competition poses an interesting problem, and we will be taking some steps to improve the experience of competing.

\n

Over the next few business days we will be:

\n
    \n
  1. Shifting from F1 to mAP, which should be easier to use / more readily available. (The new metric should be precisely the OpenImages mAP metric mentioned in the Evaluation tab.)
    2. \n
  2. Updating the 8-bit PNGs to 16-bit, which should be useful for modeling.
    3. \n
  3. Updating the public dataset to make it easier / faster to download.
    4. \n
\n

Note that when #1 and #2 are complete we will be re-scoring submissions.

\n

Thanks again for all your feedback, and please let us know if you have questions. We will post messages about updates as they occur.

", "rawMarkdown": "Hi all -\n\nWe've had some excellent early feedback in this competition - thank you to everyone participating so far! This competition poses an interesting problem, and we will be taking some steps to improve the experience of competing.\n\nOver the next few business days we will be:\n1. Shifting from F1 to mAP, which should be easier to use / more readily available. (The new metric should be precisely the [OpenImages mAP](https://www.kaggle.com/c/open-images-2019-instance-segmentation/overview/evaluation) metric mentioned in the [Evaluation](https://admin.kaggle.com/c/hpa-single-cell-image-classification/overview/evaluation) tab.)\n2. Updating the 8-bit PNGs to 16-bit, which should be useful for modeling.\n3. Updating the public dataset to make it easier / faster to download.\n\nNote that when #1 and #2 are complete we will be re-scoring submissions.\n\nThanks again for all your feedback, and please let us know if you have questions. We will post messages about updates as they occur.", "votes": 32 }, { "id": 1186602, "postDate": "2021-02-04T23:26:14.480Z", "content": "

Hi @philculliton is it possible to put training images into several folders, so images can downlaod in small batch,
\nI already fail 5 times this weeks as zip file is too big, and no way to resume download. Already try to use google cloud to download……..
\nIts very frastrating doing this…….

", "rawMarkdown": "Hi @philculliton is it possible to put training images into several folders, so images can downlaod in small batch,\nI already fail 5 times this weeks as zip file is too big, and no way to resume download. Already try to use google cloud to download........\nIts very frastrating doing this.......\n\n", "votes": 1 }, { "id": 1184216, "postDate": "2021-02-03T12:44:22.283Z", "content": "

@philculliton is it possible also to include all 4 filter channels to a single related TF example in tfrecords?

", "rawMarkdown": "@philculliton is it possible also to include all 4 filter channels to a single related TF example in tfrecords?", "votes": 1, "replies": [ { "id": 1186404, "postDate": "2021-02-04T19:58:26.470Z", "content": "

@glushko That's an excellent suggestion, thanks. I think we may not be able to restructure the TFRecords for this competition, but I'll see if it's possible, and I'll definitely keep it in mind for future competitions with this format.

", "rawMarkdown": "@glushko That's an excellent suggestion, thanks. I think we may not be able to restructure the TFRecords for this competition, but I'll see if it's possible, and I'll definitely keep it in mind for future competitions with this format.", "votes": 3 } ] }, { "id": 1198779, "postDate": "2021-02-13T09:52:13.173Z", "content": "
\n

Updating the public dataset to make it easier / faster to download.

\n
\n

@philculliton, Is there any update on this?

", "rawMarkdown": "> Updating the public dataset to make it easier / faster to download.\n\n@philculliton, Is there any update on this?", "replies": [ { "id": 1205737, "postDate": "2021-02-17T00:00:30.917Z", "content": "

@philculliton @lnhtrang Could you please just tell us if there are plans to provide the public HPA dataset in some other form? I will patiently try to download it using the current way, if it will not be provided in another, easier way to download.

", "rawMarkdown": "@philculliton @lnhtrang Could you please just tell us if there are plans to provide the public HPA dataset in some other form? I will patiently try to download it using the [current way](https://www.kaggle.com/lnhtrang/hpa-public-data-download-and-hpacellseg), if it will not be provided in another, easier way to download.", "votes": 2 }, { "id": 1220443, "postDate": "2021-02-28T02:06:07.863Z", "content": "

It was answered here.

\n
\n

Hi! Thanks for your question. Yes - this is still in progress. We'll be making a trimmed version of the public data available as a Kaggle dataset, which should be quicker to access via Kaggle notebooks (and also be a smaller download elsewhere).

\n
", "rawMarkdown": "It was answered [here](https://www.kaggle.com/c/hpa-single-cell-image-classification/discussion/217021#1219554).\n\n> Hi! Thanks for your question. Yes - this is still in progress. We'll be making a trimmed version of the public data available as a Kaggle dataset, which should be quicker to access via Kaggle notebooks (and also be a smaller download elsewhere)." } ] }, { "id": 1186516, "postDate": "2021-02-04T21:37:09.767Z", "content": "

Could you please clarify whether images in the hidden test set are 16-bit like in public one?

", "rawMarkdown": "Could you please clarify whether images in the hidden test set are 16-bit like in public one?", "replies": [ { "id": 1186534, "postDate": "2021-02-04T21:59:16.430Z", "content": "

Hi! Yes, the images in the hidden test set are 16-bit.

", "rawMarkdown": "Hi! Yes, the images in the hidden test set are 16-bit." } ] }, { "id": 1186401, "postDate": "2021-02-04T19:54:11.360Z", "content": "

Just a note to mention that we are updating the metric and starting the re-run of past submissions today, as well as releasing the 16-bit images as an additional dataset.

", "rawMarkdown": "Just a note to mention that we are updating the metric and starting the re-run of past submissions today, as well as releasing the 16-bit images as an additional dataset." } ], "comments": [ { "id": 1186602, "author_name": "gody7334", "author_url": "", "post_date": "2021-02-04T23:26:14.480000", "content": "

Hi @philculliton is it possible to put training images into several folders, so images can downlaod in small batch,
\nI already fail 5 times this weeks as zip file is too big, and no way to resume download. Already try to use google cloud to download……..
\nIts very frastrating doing this…….

", "votes": 1, "replies": [] }, { "id": 1184216, "author_name": "Roman Glushko 🦁", "author_url": "", "post_date": "2021-02-03T12:44:22.283000", "content": "

@philculliton is it possible also to include all 4 filter channels to a single related TF example in tfrecords?

", "votes": 1, "replies": [ { "id": 1186404, "author_name": "Phil Culliton", "author_url": "", "post_date": "2021-02-04T19:58:26.470000", "content": "

@glushko That's an excellent suggestion, thanks. I think we may not be able to restructure the TFRecords for this competition, but I'll see if it's possible, and I'll definitely keep it in mind for future competitions with this format.

", "votes": 3, "replies": [] } ] }, { "id": 1198779, "author_name": "KaizaburoChubachi", "author_url": "", "post_date": "2021-02-13T09:52:13.173000", "content": "
\n

Updating the public dataset to make it easier / faster to download.

\n
\n

@philculliton, Is there any update on this?

", "votes": 0, "replies": [ { "id": 1205737, "author_name": "KaizaburoChubachi", "author_url": "", "post_date": "2021-02-17T00:00:30.917000", "content": "

@philculliton @lnhtrang Could you please just tell us if there are plans to provide the public HPA dataset in some other form? I will patiently try to download it using the current way, if it will not be provided in another, easier way to download.

", "votes": 2, "replies": [] }, { "id": 1220443, "author_name": "KaizaburoChubachi", "author_url": "", "post_date": "2021-02-28T02:06:07.863000", "content": "

It was answered here.

\n
\n

Hi! Thanks for your question. Yes - this is still in progress. We'll be making a trimmed version of the public data available as a Kaggle dataset, which should be quicker to access via Kaggle notebooks (and also be a smaller download elsewhere).

\n
", "votes": 0, "replies": [] } ] }, { "id": 1186516, "author_name": "Vladislav Ostankovich", "author_url": "", "post_date": "2021-02-04T21:37:09.767000", "content": "

Could you please clarify whether images in the hidden test set are 16-bit like in public one?

", "votes": 0, "replies": [ { "id": 1186534, "author_name": "Phil Culliton", "author_url": "", "post_date": "2021-02-04T21:59:16.430000", "content": "

Hi! Yes, the images in the hidden test set are 16-bit.

", "votes": 0, "replies": [] } ] }, { "id": 1186401, "author_name": "Phil Culliton", "author_url": "", "post_date": "2021-02-04T19:54:11.360000", "content": "

Just a note to mention that we are updating the metric and starting the re-run of past submissions today, as well as releasing the 16-bit images as an additional dataset.

", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1176726": "Hi all -\n\nWe've had some excellent early feedback in this competition - thank you to everyone participating so far! This competition poses an interesting problem, and we will be taking some steps to improve the experience of competing.\n\nOver the next few business days we will be:\n1. Shifting from F1 to mAP, which should be easier to use / more readily available. (The new metric should be precisely the [OpenImages mAP](https://www.kaggle.com/c/open-images-2019-instance-segmentation/overview/evaluation) metric mentioned in the [Evaluation](https://admin.kaggle.com/c/hpa-single-cell-image-classification/overview/evaluation) tab.)\n2. Updating the 8-bit PNGs to 16-bit, which should be useful for modeling.\n3. Updating the public dataset to make it easier / faster to download.\n\nNote that when #1 and #2 are complete we will be re-scoring submissions.\n\nThanks again for all your feedback, and please let us know if you have questions. We will post messages about updates as they occur.", "1186602": "Hi @philculliton is it possible to put training images into several folders, so images can downlaod in small batch,\nI already fail 5 times this weeks as zip file is too big, and no way to resume download. Already try to use google cloud to download........\nIts very frastrating doing this.......\n\n", "1184216": "@philculliton is it possible also to include all 4 filter channels to a single related TF example in tfrecords?", "1198779": "> Updating the public dataset to make it easier / faster to download.\n\n@philculliton, Is there any update on this?", "1186516": "Could you please clarify whether images in the hidden test set are 16-bit like in public one?", "1186401": "Just a note to mention that we are updating the metric and starting the re-run of past submissions today, as well as releasing the 16-bit images as an additional dataset." } }