Welcome to the Human Protein Atlas Image Classification Challenge!
\n\nIn this competition we're predicting patterns of sub-cellular proteins from images. You'll want to familiarize yourself with the competition metric on the Evaluation page and the file descriptions on the Data tab.
\n\nThis is a multi-label competition, where several different labels might apply to a given sample. This is also a multi-image competition - each sample consists of four differently filtered images. Make sure to read the descriptions of these images in the Data tab.
\n\nIn addition to the regular prizes, this competition will have a special prize. Please check out this link to read the rules and expectations for that prize.
\n\nPlease feel free to ask your questions in this thread.
\n\nGood luck!
", "messages": [ { "id": 398300, "postDate": "2018-10-03T21:10:09.437Z", "content": "Welcome to the Human Protein Atlas Image Classification Challenge!
\n\nIn this competition we're predicting patterns of sub-cellular proteins from images. You'll want to familiarize yourself with the competition metric on the Evaluation page and the file descriptions on the Data tab.
\n\nThis is a multi-label competition, where several different labels might apply to a given sample. This is also a multi-image competition - each sample consists of four differently filtered images. Make sure to read the descriptions of these images in the Data tab.
\n\nIn addition to the regular prizes, this competition will have a special prize. Please check out this link to read the rules and expectations for that prize.
\n\nPlease feel free to ask your questions in this thread.
\n\nGood luck!
", "rawMarkdown": "Welcome to the Human Protein Atlas Image Classification Challenge!\n\nIn this competition we're predicting patterns of sub-cellular proteins from images. You'll want to familiarize yourself with the competition metric on the [Evaluation page][1] and the file descriptions on the [Data tab][2].\n\nThis is a multi-label competition, where several different labels might apply to a given sample. This is also a multi-image competition - each sample consists of four differently filtered images. Make sure to read the descriptions of these images in the [Data tab][3].\n\nIn addition to the regular prizes, this competition will have a special prize. Please check out [this link][4] to read the rules and expectations for that prize.\n\nPlease feel free to ask your questions in this thread.\n\nGood luck!\n\n\n [1]: https://www.kaggle.com/c/human-protein-atlas-image-classification#evaluation\n [2]: https://www.kaggle.com/c/human-protein-atlas-image-classification/data\n [3]: https://www.kaggle.com/c/human-protein-atlas-image-classification/data\n [4]: https://www.kaggle.com/c/human-protein-atlas-image-classification#Special-Prize-Instructions", "votes": 4 }, { "id": 432898, "postDate": "2018-12-04T12:49:23.483Z", "content": "Are kaggle or the organisers going to comment on the data leak?
\n\nhttps://www.kaggle.com/c/human-protein-atlas-image-classification/discussion/73395
", "rawMarkdown": "Are kaggle or the organisers going to comment on the data leak?\n\nhttps://www.kaggle.com/c/human-protein-atlas-image-classification/discussion/73395", "votes": 1, "replies": [ { "id": 433023, "postDate": "2018-12-04T15:18:14.947Z", "content": "Hey Mark - yep, I've answered in thread. Thanks!
", "rawMarkdown": "Hey Mark - yep, I've answered in thread. Thanks!", "votes": 2 } ] }, { "id": 404307, "postDate": "2018-10-15T14:49:25.363Z", "content": "What macro F-score human experts achieve when labeling proteins?
", "rawMarkdown": "What macro F-score human experts achieve when labeling proteins? ", "votes": 2, "replies": [ { "id": 414633, "postDate": "2018-11-03T08:30:13.647Z", "content": "In another thread one of the hosts said experts are able to achieve 0.71
", "rawMarkdown": "In another thread one of the hosts said experts are able to achieve 0.71 " } ] }, { "id": 1175381, "postDate": "2021-01-29T05:57:06.537Z", "content": "I just received an email from Kaggle stating that this competition is active (again), with an Entry Deadline of April 21, 2021. Looks like a bug…
", "rawMarkdown": "I just received an email from Kaggle stating that this competition is active (again), with an Entry Deadline of April 21, 2021. Looks like a bug...", "replies": [ { "id": 1175389, "postDate": "2021-01-29T06:08:55.263Z", "content": "Yes, i was slightly excited about the concept of getting another gv100….
", "rawMarkdown": " Yes, i was slightly excited about the concept of getting another gv100.... " } ] }, { "id": 448946, "postDate": "2019-01-02T12:06:17.340Z", "content": "Hi - this is a very beginner question -
\n\nif I compare an image where endoplasmic recticulum is the target protein (say 018dd9a2-bbc2-11e8-b2bb-ac1f6b6435d0) and the yellow filter (which is supposed to highlight the same endoplasmic recticulum), why do the image data not match? (for example if i take its difference, i dont obtain a zero matrix)
\n\nThanks!
", "rawMarkdown": "Hi - this is a very beginner question - \n\nif I compare an image where endoplasmic recticulum is the target protein (say 018dd9a2-bbc2-11e8-b2bb-ac1f6b6435d0) and the yellow filter (which is supposed to highlight the same endoplasmic recticulum), why do the image data not match? (for example if i take its difference, i dont obtain a zero matrix)\n\nThanks!" }, { "id": 443217, "postDate": "2018-12-21T08:23:40.427Z", "content": "Hello Phil,\nCould you please explain the rules concerning reproducibility of the final submission:\nShould the whole training process be completely reproducible or only the inference part?\nThank you
", "rawMarkdown": "Hello Phil,\nCould you please explain the rules concerning reproducibility of the final submission:\nShould the whole training process be completely reproducible or only the inference part?\nThank you" }, { "id": 435941, "postDate": "2018-12-09T05:32:34.803Z", "content": "For this competition, the evaluation metric is mentioned as \"macro F1 score\". There is no formula mentioned in the \"Evaluation\" description page unlike what I have seen in other competitions. I wanted to get some clarification on it, how is the macro f1 score aggregated from each classes's f1 score, is it arithmetic or harmonic mean? Since the individual f1 scores are percentages(ratios) I am assuming the aggregation is done by using harmonic mean. If that is the case do you clip any individual f1 score to some minimal value if they are close to zero, if that is the case what are they clipped to? I am asking since my leaderboard scores swing wildly even for small perturbations and I am trying to make sense of it.
", "rawMarkdown": "For this competition, the evaluation metric is mentioned as \"macro F1 score\". There is no formula mentioned in the \"Evaluation\" description page unlike what I have seen in other competitions. I wanted to get some clarification on it, how is the macro f1 score aggregated from each classes's f1 score, is it arithmetic or harmonic mean? Since the individual f1 scores are percentages(ratios) I am assuming the aggregation is done by using harmonic mean. If that is the case do you clip any individual f1 score to some minimal value if they are close to zero, if that is the case what are they clipped to? I am asking since my leaderboard scores swing wildly even for small perturbations and I am trying to make sense of it. " }, { "id": 406963, "postDate": "2018-10-20T04:59:21.777Z", "content": "Phil
\n\nThink I have read over the special prize info a couple of times and might have missed the answers to my question.
\n\nThe special prize submission is a sizeable body of work for me - if some accident happened and I ended up in a top place how soon after the deadline Jan 10 would the full special prize submission be required?
\n\nIn Emma's comments it seems she was indicating that one submission file could be our accurate script and the other could be the speed script. That seems to suggest that Jan 10 is also the deadline for the special prize submission.
", "rawMarkdown": "Phil\n\nThink I have read over the special prize info a couple of times and might have missed the answers to my question.\n\nThe special prize submission is a sizeable body of work for me - if some accident happened and I ended up in a top place how soon after the deadline Jan 10 would the full special prize submission be required?\n\nIn Emma's comments it seems she was indicating that one submission file could be our accurate script and the other could be the speed script. That seems to suggest that Jan 10 is also the deadline for the special prize submission.\n" }, { "id": 398310, "postDate": "2018-10-03T21:49:29.937Z", "content": "Competition where the prize is a Nature publication. Well that's a first :)
", "rawMarkdown": "Competition where the prize is a Nature publication. Well that's a first :)", "replies": [ { "id": 398895, "postDate": "2018-10-04T18:53:27.140Z", "content": "\n\n\nCompetition where the prize is a Nature publication. Well that's a first :)
\n
First it is. Still, let's not be casual about the journal. It is a \"Nature journal\" publication, but there is a non-negligible difference between \"Nature\" and \"Nature Methods\".
", "rawMarkdown": "> Competition where the prize is a Nature publication. Well that's a first :)\n\nFirst it is. Still, let's not be casual about the journal. It is a \"Nature journal\" publication, but there is a non-negligible difference between \"Nature\" and \"Nature Methods\".", "votes": 1 } ] }, { "id": 442473, "postDate": "2018-12-20T02:19:31.230Z", "rawMarkdown": "", "isDeleted": true }, { "id": 398713, "postDate": "2018-10-04T13:36:40.243Z", "rawMarkdown": "", "votes": 1, "isDeleted": true } ], "comments": [ { "id": 432898, "author_name": "Mark Worrall", "author_url": "", "post_date": "2018-12-04T12:49:23.483000", "content": "Are kaggle or the organisers going to comment on the data leak?
\n\nhttps://www.kaggle.com/c/human-protein-atlas-image-classification/discussion/73395
", "votes": 1, "replies": [ { "id": 433023, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-12-04T15:18:14.947000", "content": "Hey Mark - yep, I've answered in thread. Thanks!
", "votes": 2, "replies": [] } ] }, { "id": 404307, "author_name": "pi-null-mezon", "author_url": "", "post_date": "2018-10-15T14:49:25.363000", "content": "What macro F-score human experts achieve when labeling proteins?
", "votes": 2, "replies": [ { "id": 414633, "author_name": "SputKnick", "author_url": "", "post_date": "2018-11-03T08:30:13.647000", "content": "In another thread one of the hosts said experts are able to achieve 0.71
", "votes": 0, "replies": [] } ] }, { "id": 1175381, "author_name": "Charlie Turner", "author_url": "", "post_date": "2021-01-29T05:57:06.537000", "content": "I just received an email from Kaggle stating that this competition is active (again), with an Entry Deadline of April 21, 2021. Looks like a bug…
", "votes": 0, "replies": [ { "id": 1175389, "author_name": "Moshel", "author_url": "", "post_date": "2021-01-29T06:08:55.263000", "content": "Yes, i was slightly excited about the concept of getting another gv100….
", "votes": 0, "replies": [] } ] }, { "id": 448946, "author_name": "Heisenger", "author_url": "", "post_date": "2019-01-02T12:06:17.340000", "content": "Hi - this is a very beginner question -
\n\nif I compare an image where endoplasmic recticulum is the target protein (say 018dd9a2-bbc2-11e8-b2bb-ac1f6b6435d0) and the yellow filter (which is supposed to highlight the same endoplasmic recticulum), why do the image data not match? (for example if i take its difference, i dont obtain a zero matrix)
\n\nThanks!
", "votes": 0, "replies": [] }, { "id": 443217, "author_name": "Maksim Rodin", "author_url": "", "post_date": "2018-12-21T08:23:40.427000", "content": "Hello Phil,\nCould you please explain the rules concerning reproducibility of the final submission:\nShould the whole training process be completely reproducible or only the inference part?\nThank you
", "votes": 0, "replies": [] }, { "id": 435941, "author_name": "Shiv Gowda", "author_url": "", "post_date": "2018-12-09T05:32:34.803000", "content": "For this competition, the evaluation metric is mentioned as \"macro F1 score\". There is no formula mentioned in the \"Evaluation\" description page unlike what I have seen in other competitions. I wanted to get some clarification on it, how is the macro f1 score aggregated from each classes's f1 score, is it arithmetic or harmonic mean? Since the individual f1 scores are percentages(ratios) I am assuming the aggregation is done by using harmonic mean. If that is the case do you clip any individual f1 score to some minimal value if they are close to zero, if that is the case what are they clipped to? I am asking since my leaderboard scores swing wildly even for small perturbations and I am trying to make sense of it.
", "votes": 0, "replies": [] }, { "id": 406963, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2018-10-20T04:59:21.777000", "content": "Phil
\n\nThink I have read over the special prize info a couple of times and might have missed the answers to my question.
\n\nThe special prize submission is a sizeable body of work for me - if some accident happened and I ended up in a top place how soon after the deadline Jan 10 would the full special prize submission be required?
\n\nIn Emma's comments it seems she was indicating that one submission file could be our accurate script and the other could be the speed script. That seems to suggest that Jan 10 is also the deadline for the special prize submission.
", "votes": 0, "replies": [] }, { "id": 398310, "author_name": "Miha Skalic", "author_url": "", "post_date": "2018-10-03T21:49:29.937000", "content": "Competition where the prize is a Nature publication. Well that's a first :)
", "votes": 0, "replies": [ { "id": 398895, "author_name": "Tilii", "author_url": "", "post_date": "2018-10-04T18:53:27.140000", "content": "\n\n\nCompetition where the prize is a Nature publication. Well that's a first :)
\n
First it is. Still, let's not be casual about the journal. It is a \"Nature journal\" publication, but there is a non-negligible difference between \"Nature\" and \"Nature Methods\".
", "votes": 1, "replies": [] } ] }, { "id": 442473, "author_name": "", "author_url": "", "post_date": "2018-12-20T02:19:31.230000", "content": "", "votes": 0, "replies": [] }, { "id": 398713, "author_name": "", "author_url": "", "post_date": "2018-10-04T13:36:40.243000", "content": "", "votes": 1, "replies": [] } ], "raw_markdown_by_id": { "398300": "Welcome to the Human Protein Atlas Image Classification Challenge!\n\nIn this competition we're predicting patterns of sub-cellular proteins from images. You'll want to familiarize yourself with the competition metric on the [Evaluation page][1] and the file descriptions on the [Data tab][2].\n\nThis is a multi-label competition, where several different labels might apply to a given sample. This is also a multi-image competition - each sample consists of four differently filtered images. Make sure to read the descriptions of these images in the [Data tab][3].\n\nIn addition to the regular prizes, this competition will have a special prize. Please check out [this link][4] to read the rules and expectations for that prize.\n\nPlease feel free to ask your questions in this thread.\n\nGood luck!\n\n\n [1]: https://www.kaggle.com/c/human-protein-atlas-image-classification#evaluation\n [2]: https://www.kaggle.com/c/human-protein-atlas-image-classification/data\n [3]: https://www.kaggle.com/c/human-protein-atlas-image-classification/data\n [4]: https://www.kaggle.com/c/human-protein-atlas-image-classification#Special-Prize-Instructions", "432898": "Are kaggle or the organisers going to comment on the data leak?\n\nhttps://www.kaggle.com/c/human-protein-atlas-image-classification/discussion/73395", "404307": "What macro F-score human experts achieve when labeling proteins? ", "1175381": "I just received an email from Kaggle stating that this competition is active (again), with an Entry Deadline of April 21, 2021. Looks like a bug...", "448946": "Hi - this is a very beginner question - \n\nif I compare an image where endoplasmic recticulum is the target protein (say 018dd9a2-bbc2-11e8-b2bb-ac1f6b6435d0) and the yellow filter (which is supposed to highlight the same endoplasmic recticulum), why do the image data not match? (for example if i take its difference, i dont obtain a zero matrix)\n\nThanks!", "443217": "Hello Phil,\nCould you please explain the rules concerning reproducibility of the final submission:\nShould the whole training process be completely reproducible or only the inference part?\nThank you", "435941": "For this competition, the evaluation metric is mentioned as \"macro F1 score\". There is no formula mentioned in the \"Evaluation\" description page unlike what I have seen in other competitions. I wanted to get some clarification on it, how is the macro f1 score aggregated from each classes's f1 score, is it arithmetic or harmonic mean? Since the individual f1 scores are percentages(ratios) I am assuming the aggregation is done by using harmonic mean. If that is the case do you clip any individual f1 score to some minimal value if they are close to zero, if that is the case what are they clipped to? I am asking since my leaderboard scores swing wildly even for small perturbations and I am trying to make sense of it. ", "406963": "Phil\n\nThink I have read over the special prize info a couple of times and might have missed the answers to my question.\n\nThe special prize submission is a sizeable body of work for me - if some accident happened and I ended up in a top place how soon after the deadline Jan 10 would the full special prize submission be required?\n\nIn Emma's comments it seems she was indicating that one submission file could be our accurate script and the other could be the speed script. That seems to suggest that Jan 10 is also the deadline for the special prize submission.\n", "398310": "Competition where the prize is a Nature publication. Well that's a first :)", "442473": "", "398713": "" } }