Hi Kagglers!
\n\nThe MobileODT team prepared a special webinar to help you work with our data. In this video, you will find information about cervical cancer, the procedure in which the images were taken and important features you’ll find in the images. \nAny questions can be posted here and we will do our best to answer them.
\n\nGood luck in the competition!
\n\nThe MobileODT team.
\n\nhttps://www.youtube.com/watch?v=N8OGrykSgy0&feature=youtu.be
\n\nUpdate (May-15):\n More information is available on this video by professor Albert Singer:\nhttps://www.youtube.com/watch?v=FxnYU6FOLzk&t=2s
", "messages": [ { "id": 179390, "postDate": "2017-05-01T09:05:05.873Z", "content": "Hi Kagglers!
\n\nThe MobileODT team prepared a special webinar to help you work with our data. In this video, you will find information about cervical cancer, the procedure in which the images were taken and important features you’ll find in the images. \nAny questions can be posted here and we will do our best to answer them.
\n\nGood luck in the competition!
\n\nThe MobileODT team.
\n\nhttps://www.youtube.com/watch?v=N8OGrykSgy0&feature=youtu.be
\n\nUpdate (May-15):\n More information is available on this video by professor Albert Singer:\nhttps://www.youtube.com/watch?v=FxnYU6FOLzk&t=2s
", "rawMarkdown": "Hi Kagglers!\n\nThe MobileODT team prepared a special webinar to help you work with our data. In this video, you will find information about cervical cancer, the procedure in which the images were taken and important features you’ll find in the images. \nAny questions can be posted here and we will do our best to answer them.\n \nGood luck in the competition!\n\nThe MobileODT team.\n \nhttps://www.youtube.com/watch?v=N8OGrykSgy0&feature=youtu.be\n\n**Update (May-15):**\n More information is available on this video by professor Albert Singer:\nhttps://www.youtube.com/watch?v=FxnYU6FOLzk&t=2s", "votes": 16 }, { "id": 180689, "postDate": "2017-05-06T16:35:44.983Z", "content": "Hi Yau Ben-Or,
\n\nThanks for the webinar & info !
\n\nTo help us improve the model, I would like to understand how human classify some of those non-trivial cases.
\n\nI listed some examples in this kernel, would you please briefly explain how human classify these cases? based on what observations?
", "rawMarkdown": "Hi Yau Ben-Or,\n\nThanks for the webinar & info ! \n\nTo help us improve the model, I would like to understand how human classify some of those non-trivial cases.\n\nI listed some examples in [this kernel][1], would you please briefly explain how human classify these cases? based on what observations?\n\n [1]: https://www.kaggle.com/scottykwok/making-sense-out-of-some-difficult-samples/", "votes": 3, "replies": [ { "id": 182548, "postDate": "2017-05-14T14:11:27.593Z", "content": "Hi Scotty,
\n\nThank you for collecting these images.\nOur clinical advisor is double reviewing them. I will publish his comments once he finishes.
\n\nYau
", "rawMarkdown": "Hi Scotty,\n\nThank you for collecting these images.\nOur clinical advisor is double reviewing them. I will publish his comments once he finishes.\n\nYau", "votes": 1 } ] }, { "id": 179425, "postDate": "2017-05-01T13:06:47.760Z", "content": "This video was very informative.
\n\nCould you please clarify for us if the green filter is indeed a digital filter (image processing), or if it is a physical filter, or if it is a light source?
\n\nThank you!
", "rawMarkdown": "This video was very informative.\n\nCould you please clarify for us if the green filter is indeed a digital filter (image processing), or if it is a physical filter, or if it is a light source?\n\nThank you!", "votes": 3, "replies": [ { "id": 179725, "postDate": "2017-05-02T17:14:22.037Z", "content": "Hi authman,
\n\nYes, the green filter is a digital filter, basically the green channel of the RGB image.
\n\nYau
", "rawMarkdown": "Hi authman,\n\nYes, the green filter is a digital filter, basically the green channel of the RGB image.\n\nYau\n" } ] }, { "id": 179491, "postDate": "2017-05-01T17:40:23.553Z", "content": "Interesting video, thanks. I would like to ask few questions:
\n\nIs it correct that all images in the kaggle dataset that are not green filtered, with natural look like cervix (no Lugol's iodine applied) represent cervix under acetic acid application ? So, we have only three types of images: cervix under Acetic acid, cervix under Lugol's iodine and \"green\" cervix?
Could you please, explain the section of \"Edge cases\". What are exactly these posterior and anterior regions ?
Thank you
", "rawMarkdown": "Interesting video, thanks. I would like to ask few questions: \n\n- Is it correct that all images in the kaggle dataset that are not green filtered, with natural look like cervix (no Lugol's iodine applied) represent cervix under acetic acid application ? So, we have only three types of images: cervix under Acetic acid, cervix under Lugol's iodine and \"green\" cervix?\n\n- Could you please, explain the section of \"Edge cases\". What are exactly these posterior and anterior regions ?\n\nThank you", "votes": 1, "replies": [ { "id": 179724, "postDate": "2017-05-02T17:12:43.237Z", "content": "Hi vfdev,
\n\nYau
", "rawMarkdown": "Hi vfdev,\n\n\n1.\tIn addition to the three types you mentioned, some images were taken before the application of acetic acid or any other solution, and show the cervix “natural” condition.\n2.\tAnterior is the region closer to the abdomen, usually the top of the image. Posterior is the region closer to the back, usually the bottom of the image. In some cases, the transformation zone location (hence, the cervix type) is different in the posterior and anterior regions. \n\nYau\n" }, { "id": 182289, "postDate": "2017-05-13T06:11:43.043Z", "content": "180689
", "rawMarkdown": "180689" } ] }, { "id": 182287, "postDate": "2017-05-13T05:54:51.957Z", "content": "aaaaaa
", "rawMarkdown": "aaaaaa" }, { "id": 179519, "postDate": "2017-05-01T20:21:11.770Z", "content": "The video confirms an overlap between the Stage 1 test set and the additional set - can you confirm that this will not be the case with Stage 2 data? To save us having to write an image matcher as part of the solution... Thanks.
", "rawMarkdown": "The video confirms an overlap between the Stage 1 test set and the additional set - can you confirm that this will not be the case with Stage 2 data? To save us having to write an image matcher as part of the solution... Thanks.", "replies": [ { "id": 179532, "postDate": "2017-05-01T22:05:13.430Z", "content": "It was not explicitly stated that there is an overlap between test set and the additional, was it ?
", "rawMarkdown": "It was not explicitly stated that there is an overlap between test set and the additional, was it ?" }, { "id": 179600, "postDate": "2017-05-02T07:44:11.280Z", "content": "I think that is what they described from about 9:50 in the video.
", "rawMarkdown": "I think that is what they described from about 9:50 in the video." }, { "id": 179645, "postDate": "2017-05-02T11:43:10.603Z", "content": "I thought it was train set and additional, not test set and additional?
", "rawMarkdown": "I thought it was train set and additional, not test set and additional?" }, { "id": 179650, "postDate": "2017-05-02T12:13:55.800Z", "content": "\"We took the best image from each session and gave it to an expert physician for review. These high quality images are in the main train and test sets. The additional images taken in this session were annotated with the same label and are in the additional images folder. The additional images are of the same patients but have different properties like angle, filters, zoom....\"
\n\nMy understanding of this is that the additional images overlap with the current test set. You can also see this in the example test 142.jpg = additional Type 1 472.jpg, but there are many more examples.
\n\nIt would be great if admins could confirm that the same will not happen with Stage 2? Otherwise it changes the solution that we need to prepare without having any real-world benefit...
", "rawMarkdown": "\"We took the best image from each session and gave it to an expert physician for review. These high quality images are in the main train and test sets. The additional images taken in this session were annotated with the same label and are in the additional images folder. The additional images are of the same patients but have different properties like angle, filters, zoom....\"\n\nMy understanding of this is that the additional images overlap with the current test set. You can also see this in the example test 142.jpg = additional Type 1 472.jpg, but there are many more examples.\n\nIt would be great if admins could confirm that the same will not happen with Stage 2? Otherwise it changes the solution that we need to prepare without having any real-world benefit...", "votes": 4 }, { "id": 179726, "postDate": "2017-05-02T17:15:42.637Z", "content": "Hi fergusoci,
\n\nYes, in stage 2 all new images will be unique.
\n\nYau
", "rawMarkdown": "Hi fergusoci,\n\nYes, in stage 2 all new images will be unique. \n\nYau\n", "votes": 3 }, { "id": 181143, "postDate": "2017-05-08T16:27:49.330Z", "content": "Hi, will additional images for stage2 test set be available?
", "rawMarkdown": "Hi, will additional images for stage2 test set be available?" }, { "id": 181251, "postDate": "2017-05-08T23:50:23.740Z", "content": "@Dmitry,
\n\nStage 2 won't have additional images. All the images will be unique and all the patients will be unique.
", "rawMarkdown": "@Dmitry,\n\nStage 2 won't have additional images. All the images will be unique and all the patients will be unique. " } ] }, { "id": 186459, "postDate": "2017-05-28T04:59:21.617Z", "rawMarkdown": "", "isDeleted": true }, { "id": 182075, "postDate": "2017-05-12T08:32:48.900Z", "content": "Thank you for the video.
", "rawMarkdown": "Thank you for the video." } ], "comments": [ { "id": 180689, "author_name": "Scotty", "author_url": "", "post_date": "2017-05-06T16:35:44.983000", "content": "Hi Yau Ben-Or,
\n\nThanks for the webinar & info !
\n\nTo help us improve the model, I would like to understand how human classify some of those non-trivial cases.
\n\nI listed some examples in this kernel, would you please briefly explain how human classify these cases? based on what observations?
", "votes": 3, "replies": [ { "id": 182548, "author_name": "Yau Ben-Or", "author_url": "", "post_date": "2017-05-14T14:11:27.593000", "content": "Hi Scotty,
\n\nThank you for collecting these images.\nOur clinical advisor is double reviewing them. I will publish his comments once he finishes.
\n\nYau
", "votes": 1, "replies": [] } ] }, { "id": 179425, "author_name": "عثمان", "author_url": "", "post_date": "2017-05-01T13:06:47.760000", "content": "This video was very informative.
\n\nCould you please clarify for us if the green filter is indeed a digital filter (image processing), or if it is a physical filter, or if it is a light source?
\n\nThank you!
", "votes": 3, "replies": [ { "id": 179725, "author_name": "Yau Ben-Or", "author_url": "", "post_date": "2017-05-02T17:14:22.037000", "content": "Hi authman,
\n\nYes, the green filter is a digital filter, basically the green channel of the RGB image.
\n\nYau
", "votes": 0, "replies": [] } ] }, { "id": 179491, "author_name": "vfdev", "author_url": "", "post_date": "2017-05-01T17:40:23.553000", "content": "Interesting video, thanks. I would like to ask few questions:
\n\nIs it correct that all images in the kaggle dataset that are not green filtered, with natural look like cervix (no Lugol's iodine applied) represent cervix under acetic acid application ? So, we have only three types of images: cervix under Acetic acid, cervix under Lugol's iodine and \"green\" cervix?
Could you please, explain the section of \"Edge cases\". What are exactly these posterior and anterior regions ?
Thank you
", "votes": 1, "replies": [ { "id": 179724, "author_name": "Yau Ben-Or", "author_url": "", "post_date": "2017-05-02T17:12:43.237000", "content": "Hi vfdev,
\n\nYau
", "votes": 0, "replies": [] }, { "id": 182289, "author_name": "", "author_url": "", "post_date": "2017-05-13T06:11:43.043000", "content": "180689
", "votes": 0, "replies": [] } ] }, { "id": 182287, "author_name": "", "author_url": "", "post_date": "2017-05-13T05:54:51.957000", "content": "aaaaaa
", "votes": 0, "replies": [] }, { "id": 179519, "author_name": "fergusoci", "author_url": "", "post_date": "2017-05-01T20:21:11.770000", "content": "The video confirms an overlap between the Stage 1 test set and the additional set - can you confirm that this will not be the case with Stage 2 data? To save us having to write an image matcher as part of the solution... Thanks.
", "votes": 0, "replies": [ { "id": 179532, "author_name": "Alchemist", "author_url": "", "post_date": "2017-05-01T22:05:13.430000", "content": "It was not explicitly stated that there is an overlap between test set and the additional, was it ?
", "votes": 0, "replies": [] }, { "id": 179600, "author_name": "fergusoci", "author_url": "", "post_date": "2017-05-02T07:44:11.280000", "content": "I think that is what they described from about 9:50 in the video.
", "votes": 0, "replies": [] }, { "id": 179645, "author_name": "alutrin", "author_url": "", "post_date": "2017-05-02T11:43:10.603000", "content": "I thought it was train set and additional, not test set and additional?
", "votes": 0, "replies": [] }, { "id": 179650, "author_name": "fergusoci", "author_url": "", "post_date": "2017-05-02T12:13:55.800000", "content": "\"We took the best image from each session and gave it to an expert physician for review. These high quality images are in the main train and test sets. The additional images taken in this session were annotated with the same label and are in the additional images folder. The additional images are of the same patients but have different properties like angle, filters, zoom....\"
\n\nMy understanding of this is that the additional images overlap with the current test set. You can also see this in the example test 142.jpg = additional Type 1 472.jpg, but there are many more examples.
\n\nIt would be great if admins could confirm that the same will not happen with Stage 2? Otherwise it changes the solution that we need to prepare without having any real-world benefit...
", "votes": 4, "replies": [] }, { "id": 179726, "author_name": "Yau Ben-Or", "author_url": "", "post_date": "2017-05-02T17:15:42.637000", "content": "Hi fergusoci,
\n\nYes, in stage 2 all new images will be unique.
\n\nYau
", "votes": 3, "replies": [] }, { "id": 181143, "author_name": "Dmitry Ulyanov", "author_url": "", "post_date": "2017-05-08T16:27:49.330000", "content": "Hi, will additional images for stage2 test set be available?
", "votes": 0, "replies": [] }, { "id": 181251, "author_name": "Wendy Kan", "author_url": "", "post_date": "2017-05-08T23:50:23.740000", "content": "@Dmitry,
\n\nStage 2 won't have additional images. All the images will be unique and all the patients will be unique.
", "votes": 0, "replies": [] } ] }, { "id": 186459, "author_name": "", "author_url": "", "post_date": "2017-05-28T04:59:21.617000", "content": "", "votes": 0, "replies": [] }, { "id": 182075, "author_name": "VinodRai", "author_url": "", "post_date": "2017-05-12T08:32:48.900000", "content": "Thank you for the video.
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "179390": "Hi Kagglers!\n\nThe MobileODT team prepared a special webinar to help you work with our data. In this video, you will find information about cervical cancer, the procedure in which the images were taken and important features you’ll find in the images. \nAny questions can be posted here and we will do our best to answer them.\n \nGood luck in the competition!\n\nThe MobileODT team.\n \nhttps://www.youtube.com/watch?v=N8OGrykSgy0&feature=youtu.be\n\n**Update (May-15):**\n More information is available on this video by professor Albert Singer:\nhttps://www.youtube.com/watch?v=FxnYU6FOLzk&t=2s", "180689": "Hi Yau Ben-Or,\n\nThanks for the webinar & info ! \n\nTo help us improve the model, I would like to understand how human classify some of those non-trivial cases.\n\nI listed some examples in [this kernel][1], would you please briefly explain how human classify these cases? based on what observations?\n\n [1]: https://www.kaggle.com/scottykwok/making-sense-out-of-some-difficult-samples/", "179425": "This video was very informative.\n\nCould you please clarify for us if the green filter is indeed a digital filter (image processing), or if it is a physical filter, or if it is a light source?\n\nThank you!", "179491": "Interesting video, thanks. I would like to ask few questions: \n\n- Is it correct that all images in the kaggle dataset that are not green filtered, with natural look like cervix (no Lugol's iodine applied) represent cervix under acetic acid application ? So, we have only three types of images: cervix under Acetic acid, cervix under Lugol's iodine and \"green\" cervix?\n\n- Could you please, explain the section of \"Edge cases\". What are exactly these posterior and anterior regions ?\n\nThank you", "182287": "aaaaaa", "179519": "The video confirms an overlap between the Stage 1 test set and the additional set - can you confirm that this will not be the case with Stage 2 data? To save us having to write an image matcher as part of the solution... Thanks.", "186459": "", "182075": "Thank you for the video." } }