{ "id": 64270, "title": "Welcome!", "url": "/competitions/rsna-pneumonia-detection-challenge/discussion/64270", "author_name": "Phil Culliton", "post_date": "2018-08-27T16:03:29.885000", "votes": 19, "comment_count": 62, "views": 0, "content": "

Welcome to RSNA Pneumonia Detection Challenge!

\n\n

In this competition, we are finding and localizing pneumonia in images of human lungs. You should familiarize yourself with the competition metric - as well as the concept of Intersection over Union if you haven't encountered it - and the file descriptions on the Data tab.

\n\n

Also be sure to check out our excellent Getting Started resources at:

\n\n

https://www.kaggle.com/c/rsna-pneumonia-detection-challenge#Getting%20Started

\n\n

We have a kernel for exploratory data analysis, as well as a starter notebook that walks through getting the data loaded and modeled. Both provide insights on the data that competitors won't want to miss.

\n\n

Note that images can contain multiple bounding boxes. Each box is represented by a row in the training data, but (as described in the Evaluation page) your submission should have one row per image that includes all of its bounding boxes. Please check the Evaluation and Data pages for a more detailed description of how this works.

\n\n

We'll start a separate thread in a few days for domain-specific questions. In the meantime, please feel free to ask your questions in this thread.

\n\n

Best of luck to you all!

", "messages": [ { "id": 376495, "postDate": "2018-08-27T16:03:29.887Z", "content": "

Welcome to RSNA Pneumonia Detection Challenge!

\n\n

In this competition, we are finding and localizing pneumonia in images of human lungs. You should familiarize yourself with the competition metric - as well as the concept of Intersection over Union if you haven't encountered it - and the file descriptions on the Data tab.

\n\n

Also be sure to check out our excellent Getting Started resources at:

\n\n

https://www.kaggle.com/c/rsna-pneumonia-detection-challenge#Getting%20Started

\n\n

We have a kernel for exploratory data analysis, as well as a starter notebook that walks through getting the data loaded and modeled. Both provide insights on the data that competitors won't want to miss.

\n\n

Note that images can contain multiple bounding boxes. Each box is represented by a row in the training data, but (as described in the Evaluation page) your submission should have one row per image that includes all of its bounding boxes. Please check the Evaluation and Data pages for a more detailed description of how this works.

\n\n

We'll start a separate thread in a few days for domain-specific questions. In the meantime, please feel free to ask your questions in this thread.

\n\n

Best of luck to you all!

", "rawMarkdown": "Welcome to RSNA Pneumonia Detection Challenge!\n\nIn this competition, we are finding and localizing pneumonia in images of human lungs. You should familiarize yourself with the competition metric - as well as the concept of Intersection over Union if you haven't encountered it - and the file descriptions on the [Data][1] tab.\n\nAlso be sure to check out our excellent Getting Started resources at:\n\nhttps://www.kaggle.com/c/rsna-pneumonia-detection-challenge#Getting%20Started\n\nWe have a kernel for exploratory data analysis, as well as a starter notebook that walks through getting the data loaded and modeled. Both provide insights on the data that competitors won't want to miss.\n\nNote that images can contain **multiple bounding boxes**. Each box is represented by a row in the training data, but (as described in the [Evaluation][2] page) your submission should have one row per image that includes all of its bounding boxes. Please check the [Evaluation][3] and [Data][4] pages for a more detailed description of how this works.\n\nWe'll start a separate thread in a few days for domain-specific questions. In the meantime, please feel free to ask your questions in this thread.\n\nBest of luck to you all!\n\n\n [1]: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/data\n [2]: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge#evaluation\n [3]: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge#evaluation\n [4]: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/data", "votes": 19 }, { "id": 383294, "postDate": "2018-09-08T09:38:53.277Z", "content": "

although the information in the evaluation page is probably enough it would be fantastic if the host or support team will release a github or kernel with official evaluation matrix that we can run against subset of gt. I would guess that in the long run it would save us millions of submissions and millions of questions for the host team.\nA python script that takes gt format csv and predictions csv and calculate score would really really help.

", "rawMarkdown": "although the information in the evaluation page is probably enough it would be fantastic if the host or support team will release a github or kernel with official evaluation matrix that we can run against subset of gt. I would guess that in the long run it would save us millions of submissions and millions of questions for the host team.\nA python script that takes gt format csv and predictions csv and calculate score would really really help.", "votes": 3 }, { "id": 380154, "postDate": "2018-09-01T20:55:47.297Z", "content": "

The evaluation script is really needed - https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64542

", "rawMarkdown": "The evaluation script is really needed - https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64542", "votes": 3, "replies": [ { "id": 381486, "postDate": "2018-09-04T17:37:21.993Z", "content": "

Thanks Ivan. I have provided a response in the thread.

", "rawMarkdown": "Thanks Ivan. I have provided a response in the thread." } ] }, { "id": 378690, "postDate": "2018-08-30T13:29:28.320Z", "content": "

Hi @philculliton, if an image has no ground-truth bounding box, then no matter what the prediction string is (even leaving as blank), the score for that image is always 0, is that right?

", "rawMarkdown": "Hi @philculliton, if an image has no ground-truth bounding box, then no matter what the prediction string is (even leaving as blank), the score for that image is always 0, is that right? ", "votes": 1, "replies": [ { "id": 378701, "postDate": "2018-08-30T13:37:59.513Z", "content": "

If the image has no ground-truth bounding box, and the prediction is empty, the image is not included in the mean. If any bounding boxes are predicted for the image, the score is 0 and it is included in the mean.

", "rawMarkdown": "If the image has no ground-truth bounding box, and the prediction is empty, the image is not included in the mean. If any bounding boxes are predicted for the image, the score is 0 and it is included in the mean.", "votes": 5 } ] }, { "id": 377439, "postDate": "2018-08-29T06:48:19.627Z", "content": "

In the light of recent leak from \"Airbus Ship Detection Challenge\" and based on the fact that in the description is stipulated that the data source is \"NIHCC\" that contains more labeled images than the current set - Can the organizers confirm that the \"test\" data is not based on \"train\" data in NIHCC?

", "rawMarkdown": "In the light of recent leak from \"Airbus Ship Detection Challenge\" and based on the fact that in the description is stipulated that the data source is \"[NIHCC][1]\" that contains more labeled images than the current set - Can the organizers confirm that the \"test\" data is not based on \"train\" data in NIHCC?\n\n\n [1]: https://nihcc.app.box.com/v/ChestXray-NIHCC/folder/36938765345", "votes": 1, "replies": [ { "id": 377902, "postDate": "2018-08-29T23:32:36.647Z", "content": "

Hi Mihai,

\n\n

The labels used for this competition - the bounding boxes and pneumonia diagnoses - are definitely not from NIHCC. The images used here, however, are from NIHCC.

", "rawMarkdown": "Hi Mihai,\n\nThe labels used for this competition - the bounding boxes and pneumonia diagnoses - are definitely not from NIHCC. The images used here, however, are from NIHCC.", "votes": 1 }, { "id": 378678, "postDate": "2018-08-30T13:21:03.780Z", "content": "

I've sent you a private message .

", "rawMarkdown": "I've sent you a private message ." }, { "id": 378888, "postDate": "2018-08-30T16:33:07.577Z", "content": "

Hi Mihai,

\n\n

Thanks very much for your message. Please see our response in this thread regarding using external annotations like these.

", "rawMarkdown": "Hi Mihai,\n\nThanks very much for your message. Please see our response in [this thread][1] regarding using external annotations like these.\n\n\n [1]: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64345#378885" } ] }, { "id": 376739, "postDate": "2018-08-28T02:08:06.097Z", "content": "

Hi Phil, I have two questions: \ncould you explain the \"confidence\" number that should be submitted with each bounding box. should it be between 0 and 1?\nAlso, in the rules I see that the \"model\" is required to be submitted by the end of stage 1. Is \"model\" means the source code? Does it matter what programming software we are using?

\n\n

thanks.

", "rawMarkdown": "Hi Phil, I have two questions: \ncould you explain the \"confidence\" number that should be submitted with each bounding box. should it be between 0 and 1?\nAlso, in the rules I see that the \"model\" is required to be submitted by the end of stage 1. Is \"model\" means the source code? Does it matter what programming software we are using?\n\nthanks.\n ", "votes": 1, "replies": [ { "id": 376816, "postDate": "2018-08-28T05:31:48.587Z", "content": "

I am also confused about 'confidence'

", "rawMarkdown": "I am also confused about 'confidence' " }, { "id": 377149, "postDate": "2018-08-28T16:40:09.030Z", "content": "

Hey Rami -

\n\n

The confidence number should ideally be between 0 and 1. It can be any value, but 0 and 1 are useful boundaries if you want to generate those values using a model (say, using a probability as a confidence score).

\n\n

For the \"model\" - it needs to be possible to verify that the model you upload at the end of stage 1 is the same one that was used to generate your stage 2 submission. So source code and any other supporting documentation you think would be necessary to do that should be submitted.

\n\n

Some guidelines are provided on the Timeline page.

", "rawMarkdown": "Hey Rami - \n\nThe confidence number should ideally be between 0 and 1. It can be any value, but 0 and 1 are useful boundaries if you want to generate those values using a model (say, using a probability as a confidence score).\n\nFor the \"model\" - it needs to be possible to verify that the model you upload at the end of stage 1 is the same one that was used to generate your stage 2 submission. So source code and any other supporting documentation you think would be necessary to do that should be submitted.\n\nSome guidelines are provided on the [Timeline][1] page.\n\n [1]: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge#Timeline" }, { "id": 377154, "postDate": "2018-08-28T16:45:45.663Z", "content": "

Hey Attaullah -

\n\n

I'm happy to help - what can I clear up about confidence?

", "rawMarkdown": "Hey Attaullah -\n\nI'm happy to help - what can I clear up about confidence?" }, { "id": 377306, "postDate": "2018-08-28T23:44:03.047Z", "content": "

Thanks Phil. Your previous reply to Rami, cleared my misconception.

", "rawMarkdown": "Thanks Phil. Your previous reply to Rami, cleared my misconception." }, { "id": 377483, "postDate": "2018-08-29T08:32:22.067Z", "content": "

An additional question about role of confidence in evaluation： At some threshold t, it seems that all three cases(TP, FP or FN) of bounding box are evaluated by its IoU and IoU only, how does confidence affect the score?

", "rawMarkdown": "An additional question about role of confidence in [evaluation][1]： At some threshold t, it seems that all three cases(TP, FP or FN) of bounding box are evaluated by its IoU and IoU only, how does confidence affect the score?\n\n [1]: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge#evaluation" }, { "id": 381489, "postDate": "2018-09-04T17:38:24.873Z", "content": "

Thanks Firebomber. I have updated the Evaluation page with more information, and have provided a general response to questions about confidence in this thread.

", "rawMarkdown": "Thanks Firebomber. I have updated the Evaluation page with more information, and have provided a general response to questions about `confidence` in [this thread][1].\n\n\n [1]: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64860" } ] }, { "id": 391109, "postDate": "2018-09-21T09:21:13.987Z", "content": "

can you please address https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/66323 ?

", "rawMarkdown": "can you please address https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/66323 ?", "votes": 2 }, { "id": 1121534, "postDate": "2020-12-21T18:11:35.473Z", "content": "

What type of preprocessing should I do for those X-ray Images and what is best technique to find the opacity ?
\nPlease anyone help me regarding this :)

", "rawMarkdown": "What type of preprocessing should I do for those X-ray Images and what is best technique to find the opacity ? \nPlease anyone help me regarding this :)" }, { "id": 755151, "postDate": "2020-02-24T13:48:35.877Z", "content": "

Great dataset! I'm very interested in this competition!

", "rawMarkdown": "Great dataset! I'm very interested in this competition!" }, { "id": 525984, "postDate": "2019-05-02T05:51:15.313Z", "content": "

How can i download dataset?

", "rawMarkdown": "How can i download dataset?\n" }, { "id": 405712, "postDate": "2018-10-18T01:12:30.813Z", "content": "

our team missed the team merging deadline... we were too busy with developing the model. Is there any way for us to ask for a team merge privately? Hope to get a reply. Thank you

", "rawMarkdown": "our team missed the team merging deadline... we were too busy with developing the model. Is there any way for us to ask for a team merge privately? Hope to get a reply. Thank you" }, { "id": 405342, "postDate": "2018-10-17T09:20:37.557Z", "content": "

I have a question about the evaluation metric. If we predict a box with IoU below the threshold, would the box be counted as a false negative or a false positive? What if we also predicted another box that had better IoU with the same ground truth box? Would it then be counted as best of the 2 or as 1 true positive and 1 false positive?

", "rawMarkdown": "I have a question about the evaluation metric. If we predict a box with IoU below the threshold, would the box be counted as a false negative or a false positive? What if we also predicted another box that had better IoU with the same ground truth box? Would it then be counted as best of the 2 or as 1 true positive and 1 false positive?", "replies": [ { "id": 405651, "postDate": "2018-10-17T21:18:24.113Z", "content": "

Good question! If you predict a box with IoU below the threshold for all ground truth boxes, the box is counted as a false positive. If a predicted box has an IoU above the threshold, it's counted as a true positive. So in the case you're talking about, you'd have 1 false positive and 1 true positive.

", "rawMarkdown": "Good question! If you predict a box with IoU below the threshold for all ground truth boxes, the box is counted as a false positive. If a predicted box has an IoU above the threshold, it's counted as a true positive. So in the case you're talking about, you'd have 1 false positive and 1 true positive.", "votes": 1 }, { "id": 405895, "postDate": "2018-10-18T09:50:34.583Z", "content": "

Thanks!

", "rawMarkdown": "Thanks!" } ] }, { "id": 404355, "postDate": "2018-10-15T16:10:19.683Z", "content": "

Great dataset! An awesome competition for those who want to start applying DL for medical data.

", "rawMarkdown": "Great dataset! An awesome competition for those who want to start applying DL for medical data. " }, { "id": 390999, "postDate": "2018-09-21T06:10:28.307Z", "content": "

writing it just to lift my rank....sorry

", "rawMarkdown": "writing it just to lift my rank....sorry" }, { "id": 381648, "postDate": "2018-09-04T23:29:09.247Z", "content": "

Great dataset! I'm looking forward to joining the competition!

", "rawMarkdown": "Great dataset! I'm looking forward to joining the competition!" }, { "id": 380626, "postDate": "2018-09-03T05:40:25.600Z", "content": "

I know, i shouldn't ask this , but is it necessary to download this 3gb of data or any other option is available .

", "rawMarkdown": "I know, i shouldn't ask this , but is it necessary to download this 3gb of data or any other option is available .", "replies": [ { "id": 381650, "postDate": "2018-09-04T23:30:45.650Z", "content": "

It looks as though you can add the dataset within a Kernel without having to download.

", "rawMarkdown": "It looks as though you can add the dataset within a Kernel without having to download. ", "votes": 1 } ] }, { "id": 380142, "postDate": "2018-09-01T19:53:47.650Z", "content": "

Could you please clarify on this - https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64644\n?

", "rawMarkdown": "Could you please clarify on this - https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64644\n?" }, { "id": 379210, "postDate": "2018-08-31T02:49:27.480Z", "content": "

Hi Phil,\nWould you please confirm the number of training images in stage_1_train_images.zip? After unzipping, I'm showing there are 25,685 dcm files while the stage_1_train_labels.csv file has 28,989 patient entries. Thanks for a great contest!

", "rawMarkdown": "Hi Phil,\nWould you please confirm the number of training images in stage_1_train_images.zip? After unzipping, I'm showing there are 25,685 dcm files while the stage_1_train_labels.csv file has 28,989 patient entries. Thanks for a great contest!", "replies": [ { "id": 379644, "postDate": "2018-08-31T17:40:15.110Z", "content": "

Hi David,

\n\n

Some images have multiple bounding boxes. Each bounding box is a row in the training labels file. There are the same number of unique patientIds in the training set as there are images.

", "rawMarkdown": "Hi David,\n\nSome images have multiple bounding boxes. Each bounding box is a row in the training labels file. There are the same number of unique patientIds in the training set as there are images." } ] }, { "id": 379051, "postDate": "2018-08-30T19:34:48.070Z", "content": "

I'm wondering about the GCP credits. Under \"Getting Started\" it says \"View the Data page for the URL\" but I see no URL there. Is this because:

\n\n
    \n
  1. All credits have already been awarded, and the URL has been removed;
    2. \n
  2. The URL hasn't been added yet;
    3. \n
  3. It is only visible to people who have already made a submission; or
    4. \n
  4. I didn't look closely enough?
    5. \n
\n\n

(Note: I have accepted the rules but have not made a submission. Have not decided whether to do so yet, but the \"get 'em before they run out\" offer would certainly be an issue in that decision.)

", "rawMarkdown": "I'm wondering about the GCP credits. Under \"Getting Started\" it says \"View the Data page for the URL\" but I see no URL there. Is this because:\n\n 1. All credits have already been awarded, and the URL has been removed;\n 2. The URL hasn't been added yet; \n 3. It is only visible to people who have already made a submission; or\n 4. I didn't look closely enough?\n\n(Note: I have accepted the rules but have not made a submission. Have not decided whether to do so yet, but the \"get 'em before they run out\" offer would certainly be an issue in that decision.)", "replies": [ { "id": 379062, "postDate": "2018-08-30T19:52:31.140Z", "content": "

The GCP credit URL is in a separate file in the competition data, called \"GCP Credits Request Link - RSNA.txt\".

", "rawMarkdown": "The GCP credit URL is in a separate file in the competition data, called \"GCP Credits Request Link - RSNA.txt\"." }, { "id": 379063, "postDate": "2018-08-30T19:55:00.007Z", "content": "

Yes, sorry, just now I was looking at the data and realized that #4 on my list of possibilities was the right one :)

", "rawMarkdown": "Yes, sorry, just now I was looking at the data and realized that #4 on my list of possibilities was the right one :)" }, { "id": 379068, "postDate": "2018-08-30T20:03:36.823Z", "content": "

How would mergers affect the credits? For example, if two teams that have each received credits subsequently merge, are both credits still valid?

", "rawMarkdown": "How would mergers affect the credits? For example, if two teams that have each received credits subsequently merge, are both credits still valid?" }, { "id": 379077, "postDate": "2018-08-30T20:24:01.640Z", "content": "

Yes. Both would still be valid.

", "rawMarkdown": "Yes. Both would still be valid." }, { "id": 379233, "postDate": "2018-08-31T03:55:49.647Z", "content": "

When will the credits be given? I am waiting for the credits to get started on the competition

", "rawMarkdown": "When will the credits be given? I am waiting for the credits to get started on the competition" }, { "id": 382327, "postDate": "2018-09-06T06:18:11.600Z", "content": "

@Phil : When I am opening the GCP credit url provided in the \"GCP Credits Request Link - RSNA.txt\" file, it says \"404 page not found\". Any update on it?

", "rawMarkdown": "@Phil : When I am opening the GCP credit url provided in the \"GCP Credits Request Link - RSNA.txt\" file, it says \"404 page not found\". Any update on it?" }, { "id": 383104, "postDate": "2018-09-07T17:18:29.237Z", "content": "

We have hit the maximum number of coupons to issue, so the form has been disabled. For those who have submitted eligible requests, you should receive your coupons soon.

", "rawMarkdown": "We have hit the maximum number of coupons to issue, so the form has been disabled. For those who have submitted eligible requests, you should receive your coupons soon.", "votes": 1 } ] }, { "id": 378682, "postDate": "2018-08-30T13:24:35.693Z", "content": "

Hi Phil. Should I upload (and license OpenSource if I win) my training code, as well as inference code? Or, if my uploaded model and inference code can reproduce the result, do I only have to upload them? Thank you!

", "rawMarkdown": "Hi Phil. Should I upload (and license OpenSource if I win) my training code, as well as inference code? Or, if my uploaded model and inference code can reproduce the result, do I only have to upload them? Thank you!", "replies": [ { "id": 379131, "postDate": "2018-08-30T23:19:13.593Z", "content": "

Hi Yuchiro,

\n\n

Sure! There's a great explanation of what we're looking for in the model upload process here. The important thing is that it needs to be possible to verify that the model you upload at the end of stage 1 is the same one that was used to generate your stage 2 submission. So source code and any other supporting documentation that you think would be necessary to do that should be submitted.

", "rawMarkdown": "Hi Yuchiro,\n\nSure! There's a great explanation of what we're looking for in the model upload process [here][1]. The important thing is that it needs to be possible to verify that the model you upload at the end of stage 1 is the same one that was used to generate your stage 2 submission. So source code and any other supporting documentation that you think would be necessary to do that should be submitted.\n\n\n [1]: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64336#376968", "votes": 1 } ] }, { "id": 376714, "postDate": "2018-08-28T01:09:38.033Z", "content": "

Hi, Phill!\nMay I confirm about the label data? (stage_1_train_labels.csv)

\n\n

I found some duplicated patient ID in labels. \nThat is what you mentioned about the multiple bouncing box, doesn't it? And Does it just mean the label is stored in multiple rows in train_labels, doesn't it?\n(For example, patientId = '7d674c82-5501-4730-92c5-d241fd6911e7' appears 4 times in labels.csv file.)

\n\n

Thanks.

", "rawMarkdown": "Hi, Phill!\nMay I confirm about the label data? (stage_1_train_labels.csv)\n\nI found some duplicated patient ID in labels. \nThat is what you mentioned about the multiple bouncing box, doesn't it? And Does it just mean the label is stored in multiple rows in train_labels, doesn't it?\n(For example, patientId = '7d674c82-5501-4730-92c5-d241fd6911e7' appears 4 times in labels.csv file.)\n\nThanks.", "replies": [ { "id": 376719, "postDate": "2018-08-28T01:19:53.200Z", "content": "

That's correct! The patient ID has multiple bounding boxes, provided over multiple rows.

", "rawMarkdown": "That's correct! The patient ID has multiple bounding boxes, provided over multiple rows.", "votes": 2 }, { "id": 376728, "postDate": "2018-08-28T01:42:51.217Z", "content": "

Thanks! Now I'm clear that the label data and submission file format is different.

", "rawMarkdown": "Thanks! Now I'm clear that the label data and submission file format is different.", "votes": 1 }, { "id": 385696, "postDate": "2018-09-11T13:14:11.210Z", "content": "

Hey masato and phil. I want your help in understanding the bounding boxes representation. We know that more than 1 row in training label file represents more than one bounding boxes. So how the dimensions of each bounding box will be interpreted. I mean that only one coordinate of X & Y is given and I am not able to understand that whether it is top left, right or bottom left, right coordinate. I hope my question is clear and making sense to you guys. Please solve my dilemma. Thanks in advance.

", "rawMarkdown": "Hey masato and phil. I want your help in understanding the bounding boxes representation. We know that more than 1 row in training label file represents more than one bounding boxes. So how the dimensions of each bounding box will be interpreted. I mean that only one coordinate of X & Y is given and I am not able to understand that whether it is top left, right or bottom left, right coordinate. I hope my question is clear and making sense to you guys. Please solve my dilemma. Thanks in advance. " } ] }, { "id": 376611, "postDate": "2018-08-27T19:41:31.207Z", "content": "

Thanks Phil. I am unable to download the data. I get a network error if I try downloading it directly. Or a \"TqdmKeyError(\"Unknown argument(s): \" + str(kwargs)))\" error if I use the kaggle api. Is there a problem at your end or am I downloading it incorrectly.

", "rawMarkdown": "Thanks Phil. I am unable to download the data. I get a network error if I try downloading it directly. Or a \"TqdmKeyError(\"Unknown argument(s): \" + str(kwargs)))\" error if I use the kaggle api. Is there a problem at your end or am I downloading it incorrectly.", "replies": [ { "id": 376616, "postDate": "2018-08-27T19:50:39.067Z", "content": "

Thanks. We'll check it out.

", "rawMarkdown": "Thanks. We'll check it out." }, { "id": 376618, "postDate": "2018-08-27T19:55:11Z", "content": "

Which version of the API are you using?

", "rawMarkdown": "Which version of the API are you using?" }, { "id": 376619, "postDate": "2018-08-27T19:58:38.763Z", "content": "

Sorry for the trouble. Was able to download it on the 4th try.

", "rawMarkdown": "Sorry for the trouble. Was able to download it on the 4th try." }, { "id": 376626, "postDate": "2018-08-27T20:46:44.743Z", "content": "

I'm glad you resolved the problem.

", "rawMarkdown": "I'm glad you resolved the problem." } ] }, { "id": 376568, "postDate": "2018-08-27T18:30:12.730Z", "content": "

Thanks Phil. The link to the starter kernel is bad - 404 error.

", "rawMarkdown": "Thanks Phil. The link to the starter kernel is bad - 404 error.", "replies": [ { "id": 376576, "postDate": "2018-08-27T18:39:52.643Z", "content": "

Thanks Dan! Should be available now.

", "rawMarkdown": "Thanks Dan! Should be available now." }, { "id": 379137, "postDate": "2018-08-30T23:38:06.780Z", "content": "

What is the link to the starter code?

", "rawMarkdown": "What is the link to the starter code?" }, { "id": 381491, "postDate": "2018-09-04T17:39:55.960Z", "content": "

The page containing the starter code is here. It is also linked in the original post at the top of this thread.

", "rawMarkdown": "The page containing the starter code is [here][1]. It is also linked in the original post at the top of this thread.\n\n\n [1]: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge#evaluation" }, { "id": 381585, "postDate": "2018-09-04T20:11:25.833Z", "content": "

Thanks.

", "rawMarkdown": "Thanks." } ] }, { "id": 399807, "postDate": "2018-10-06T20:13:24.633Z", "rawMarkdown": "", "isDeleted": true }, { "id": 379208, "postDate": "2018-08-31T02:45:07.080Z", "rawMarkdown": "", "isDeleted": true, "replies": [ { "id": 379211, "postDate": "2018-08-31T02:50:33.140Z", "content": "

Some images have multiple bounding boxes. Each bounding box is a row. There are the same number of unique patientIds in the training set as there are images.

", "rawMarkdown": "Some images have multiple bounding boxes. Each bounding box is a row. There are the same number of unique patientIds in the training set as there are images.", "votes": 1 } ] }, { "id": 377069, "postDate": "2018-08-28T14:45:03.960Z", "rawMarkdown": "", "isDeleted": true }, { "id": 377174, "postDate": "2018-08-28T17:23:01.297Z", "content": "

Thank you Phil.

", "rawMarkdown": "Thank you Phil." }, { "id": 376823, "postDate": "2018-08-28T05:49:54.280Z", "content": "

Thanks Phil.

", "rawMarkdown": "Thanks Phil." } ], "comments": [ { "id": 383294, "author_name": "Moshel", "author_url": "", "post_date": "2018-09-08T09:38:53.277000", "content": "

although the information in the evaluation page is probably enough it would be fantastic if the host or support team will release a github or kernel with official evaluation matrix that we can run against subset of gt. I would guess that in the long run it would save us millions of submissions and millions of questions for the host team.\nA python script that takes gt format csv and predictions csv and calculate score would really really help.

", "votes": 3, "replies": [] }, { "id": 380154, "author_name": "Ivan Pozdnyakov", "author_url": "", "post_date": "2018-09-01T20:55:47.297000", "content": "

The evaluation script is really needed - https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64542

", "votes": 3, "replies": [ { "id": 381486, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-09-04T17:37:21.993000", "content": "

Thanks Ivan. I have provided a response in the thread.

", "votes": 0, "replies": [] } ] }, { "id": 378690, "author_name": "Kha Vo", "author_url": "", "post_date": "2018-08-30T13:29:28.320000", "content": "

Hi @philculliton, if an image has no ground-truth bounding box, then no matter what the prediction string is (even leaving as blank), the score for that image is always 0, is that right?

", "votes": 1, "replies": [ { "id": 378701, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-08-30T13:37:59.513000", "content": "

If the image has no ground-truth bounding box, and the prediction is empty, the image is not included in the mean. If any bounding boxes are predicted for the image, the score is 0 and it is included in the mean.

", "votes": 5, "replies": [] } ] }, { "id": 377439, "author_name": "Mihai Cvasnievschi", "author_url": "", "post_date": "2018-08-29T06:48:19.627000", "content": "

In the light of recent leak from \"Airbus Ship Detection Challenge\" and based on the fact that in the description is stipulated that the data source is \"NIHCC\" that contains more labeled images than the current set - Can the organizers confirm that the \"test\" data is not based on \"train\" data in NIHCC?

", "votes": 1, "replies": [ { "id": 377902, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-08-29T23:32:36.647000", "content": "

Hi Mihai,

\n\n

The labels used for this competition - the bounding boxes and pneumonia diagnoses - are definitely not from NIHCC. The images used here, however, are from NIHCC.

", "votes": 1, "replies": [] }, { "id": 378678, "author_name": "Mihai Cvasnievschi", "author_url": "", "post_date": "2018-08-30T13:21:03.780000", "content": "

I've sent you a private message .

", "votes": 0, "replies": [] }, { "id": 378888, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-08-30T16:33:07.577000", "content": "

Hi Mihai,

\n\n

Thanks very much for your message. Please see our response in this thread regarding using external annotations like these.

", "votes": 0, "replies": [] } ] }, { "id": 376739, "author_name": "Rami", "author_url": "", "post_date": "2018-08-28T02:08:06.097000", "content": "

Hi Phil, I have two questions: \ncould you explain the \"confidence\" number that should be submitted with each bounding box. should it be between 0 and 1?\nAlso, in the rules I see that the \"model\" is required to be submitted by the end of stage 1. Is \"model\" means the source code? Does it matter what programming software we are using?

\n\n

thanks.

", "votes": 1, "replies": [ { "id": 376816, "author_name": "Attaullah Sahito", "author_url": "", "post_date": "2018-08-28T05:31:48.587000", "content": "

I am also confused about 'confidence'

", "votes": 0, "replies": [] }, { "id": 377149, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-08-28T16:40:09.030000", "content": "

Hey Rami -

\n\n

The confidence number should ideally be between 0 and 1. It can be any value, but 0 and 1 are useful boundaries if you want to generate those values using a model (say, using a probability as a confidence score).

\n\n

For the \"model\" - it needs to be possible to verify that the model you upload at the end of stage 1 is the same one that was used to generate your stage 2 submission. So source code and any other supporting documentation you think would be necessary to do that should be submitted.

\n\n

Some guidelines are provided on the Timeline page.

", "votes": 0, "replies": [] }, { "id": 377154, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-08-28T16:45:45.663000", "content": "

Hey Attaullah -

\n\n

I'm happy to help - what can I clear up about confidence?

", "votes": 0, "replies": [] }, { "id": 377306, "author_name": "Attaullah Sahito", "author_url": "", "post_date": "2018-08-28T23:44:03.047000", "content": "

Thanks Phil. Your previous reply to Rami, cleared my misconception.

", "votes": 0, "replies": [] }, { "id": 377483, "author_name": "Firebomber", "author_url": "", "post_date": "2018-08-29T08:32:22.067000", "content": "

An additional question about role of confidence in evaluation： At some threshold t, it seems that all three cases(TP, FP or FN) of bounding box are evaluated by its IoU and IoU only, how does confidence affect the score?

", "votes": 0, "replies": [] }, { "id": 381489, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-09-04T17:38:24.873000", "content": "

Thanks Firebomber. I have updated the Evaluation page with more information, and have provided a general response to questions about confidence in this thread.

", "votes": 0, "replies": [] } ] }, { "id": 391109, "author_name": "Moshel", "author_url": "", "post_date": "2018-09-21T09:21:13.987000", "content": "

can you please address https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/66323 ?

", "votes": 2, "replies": [] }, { "id": 1121534, "author_name": "Rohit Wagh", "author_url": "", "post_date": "2020-12-21T18:11:35.473000", "content": "

What type of preprocessing should I do for those X-ray Images and what is best technique to find the opacity ?
\nPlease anyone help me regarding this :)

", "votes": 0, "replies": [] }, { "id": 755151, "author_name": "", "author_url": "", "post_date": "2020-02-24T13:48:35.877000", "content": "

Great dataset! I'm very interested in this competition!

", "votes": 0, "replies": [] }, { "id": 525984, "author_name": "Dr. Syed Adnan", "author_url": "", "post_date": "2019-05-02T05:51:15.313000", "content": "

How can i download dataset?

", "votes": 0, "replies": [] }, { "id": 405712, "author_name": "chadrick kwag", "author_url": "", "post_date": "2018-10-18T01:12:30.813000", "content": "

our team missed the team merging deadline... we were too busy with developing the model. Is there any way for us to ask for a team merge privately? Hope to get a reply. Thank you

", "votes": 0, "replies": [] }, { "id": 405342, "author_name": "Adithya Seshasayee", "author_url": "", "post_date": "2018-10-17T09:20:37.557000", "content": "

I have a question about the evaluation metric. If we predict a box with IoU below the threshold, would the box be counted as a false negative or a false positive? What if we also predicted another box that had better IoU with the same ground truth box? Would it then be counted as best of the 2 or as 1 true positive and 1 false positive?

", "votes": 0, "replies": [ { "id": 405651, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-10-17T21:18:24.113000", "content": "

Good question! If you predict a box with IoU below the threshold for all ground truth boxes, the box is counted as a false positive. If a predicted box has an IoU above the threshold, it's counted as a true positive. So in the case you're talking about, you'd have 1 false positive and 1 true positive.

", "votes": 1, "replies": [] }, { "id": 405895, "author_name": "Adithya Seshasayee", "author_url": "", "post_date": "2018-10-18T09:50:34.583000", "content": "

Thanks!

", "votes": 0, "replies": [] } ] }, { "id": 404355, "author_name": "Angelo Menezes", "author_url": "", "post_date": "2018-10-15T16:10:19.683000", "content": "

Great dataset! An awesome competition for those who want to start applying DL for medical data.

", "votes": 0, "replies": [] }, { "id": 390999, "author_name": "Andrew", "author_url": "", "post_date": "2018-09-21T06:10:28.307000", "content": "

writing it just to lift my rank....sorry

", "votes": 0, "replies": [] }, { "id": 381648, "author_name": "Aaron Gudmundson", "author_url": "", "post_date": "2018-09-04T23:29:09.247000", "content": "

Great dataset! I'm looking forward to joining the competition!

", "votes": 0, "replies": [] }, { "id": 380626, "author_name": "Neel kamal", "author_url": "", "post_date": "2018-09-03T05:40:25.600000", "content": "

I know, i shouldn't ask this , but is it necessary to download this 3gb of data or any other option is available .

", "votes": 0, "replies": [ { "id": 381650, "author_name": "Aaron Gudmundson", "author_url": "", "post_date": "2018-09-04T23:30:45.650000", "content": "

It looks as though you can add the dataset within a Kernel without having to download.

", "votes": 1, "replies": [] } ] }, { "id": 380142, "author_name": "Ivan Pozdnyakov", "author_url": "", "post_date": "2018-09-01T19:53:47.650000", "content": "

Could you please clarify on this - https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64644\n?

", "votes": 0, "replies": [] }, { "id": 379210, "author_name": "David Martinez", "author_url": "", "post_date": "2018-08-31T02:49:27.480000", "content": "

Hi Phil,\nWould you please confirm the number of training images in stage_1_train_images.zip? After unzipping, I'm showing there are 25,685 dcm files while the stage_1_train_labels.csv file has 28,989 patient entries. Thanks for a great contest!

", "votes": 0, "replies": [ { "id": 379644, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-08-31T17:40:15.110000", "content": "

Hi David,

\n\n

Some images have multiple bounding boxes. Each bounding box is a row in the training labels file. There are the same number of unique patientIds in the training set as there are images.

", "votes": 0, "replies": [] } ] }, { "id": 379051, "author_name": "Andy Harless", "author_url": "", "post_date": "2018-08-30T19:34:48.070000", "content": "

I'm wondering about the GCP credits. Under \"Getting Started\" it says \"View the Data page for the URL\" but I see no URL there. Is this because:

\n\n
    \n
  1. All credits have already been awarded, and the URL has been removed;
    2. \n
  2. The URL hasn't been added yet;
    3. \n
  3. It is only visible to people who have already made a submission; or
    4. \n
  4. I didn't look closely enough?
    5. \n
\n\n

(Note: I have accepted the rules but have not made a submission. Have not decided whether to do so yet, but the \"get 'em before they run out\" offer would certainly be an issue in that decision.)

", "votes": 0, "replies": [ { "id": 379062, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-08-30T19:52:31.140000", "content": "

The GCP credit URL is in a separate file in the competition data, called \"GCP Credits Request Link - RSNA.txt\".

", "votes": 0, "replies": [] }, { "id": 379063, "author_name": "Andy Harless", "author_url": "", "post_date": "2018-08-30T19:55:00.007000", "content": "

Yes, sorry, just now I was looking at the data and realized that #4 on my list of possibilities was the right one :)

", "votes": 0, "replies": [] }, { "id": 379068, "author_name": "Andy Harless", "author_url": "", "post_date": "2018-08-30T20:03:36.823000", "content": "

How would mergers affect the credits? For example, if two teams that have each received credits subsequently merge, are both credits still valid?

", "votes": 0, "replies": [] }, { "id": 379077, "author_name": "Julia Elliott", "author_url": "", "post_date": "2018-08-30T20:24:01.640000", "content": "

Yes. Both would still be valid.

", "votes": 0, "replies": [] }, { "id": 379233, "author_name": "Jithu R Jacob", "author_url": "", "post_date": "2018-08-31T03:55:49.647000", "content": "

When will the credits be given? I am waiting for the credits to get started on the competition

", "votes": 0, "replies": [] }, { "id": 382327, "author_name": "Pankaj Kumar Sharma", "author_url": "", "post_date": "2018-09-06T06:18:11.600000", "content": "

@Phil : When I am opening the GCP credit url provided in the \"GCP Credits Request Link - RSNA.txt\" file, it says \"404 page not found\". Any update on it?

", "votes": 0, "replies": [] }, { "id": 383104, "author_name": "Julia Elliott", "author_url": "", "post_date": "2018-09-07T17:18:29.237000", "content": "

We have hit the maximum number of coupons to issue, so the form has been disabled. For those who have submitted eligible requests, you should receive your coupons soon.

", "votes": 1, "replies": [] } ] }, { "id": 378682, "author_name": "Yuichiro Hirano", "author_url": "", "post_date": "2018-08-30T13:24:35.693000", "content": "

Hi Phil. Should I upload (and license OpenSource if I win) my training code, as well as inference code? Or, if my uploaded model and inference code can reproduce the result, do I only have to upload them? Thank you!

", "votes": 0, "replies": [ { "id": 379131, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-08-30T23:19:13.593000", "content": "

Hi Yuchiro,

\n\n

Sure! There's a great explanation of what we're looking for in the model upload process here. The important thing is that it needs to be possible to verify that the model you upload at the end of stage 1 is the same one that was used to generate your stage 2 submission. So source code and any other supporting documentation that you think would be necessary to do that should be submitted.

", "votes": 1, "replies": [] } ] }, { "id": 376714, "author_name": "Masato", "author_url": "", "post_date": "2018-08-28T01:09:38.033000", "content": "

Hi, Phill!\nMay I confirm about the label data? (stage_1_train_labels.csv)

\n\n

I found some duplicated patient ID in labels. \nThat is what you mentioned about the multiple bouncing box, doesn't it? And Does it just mean the label is stored in multiple rows in train_labels, doesn't it?\n(For example, patientId = '7d674c82-5501-4730-92c5-d241fd6911e7' appears 4 times in labels.csv file.)

\n\n

Thanks.

", "votes": 0, "replies": [ { "id": 376719, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-08-28T01:19:53.200000", "content": "

That's correct! The patient ID has multiple bounding boxes, provided over multiple rows.

", "votes": 2, "replies": [] }, { "id": 376728, "author_name": "Masato", "author_url": "", "post_date": "2018-08-28T01:42:51.217000", "content": "

Thanks! Now I'm clear that the label data and submission file format is different.

", "votes": 1, "replies": [] }, { "id": 385696, "author_name": "Divyang Deep Tiwari", "author_url": "", "post_date": "2018-09-11T13:14:11.210000", "content": "

Hey masato and phil. I want your help in understanding the bounding boxes representation. We know that more than 1 row in training label file represents more than one bounding boxes. So how the dimensions of each bounding box will be interpreted. I mean that only one coordinate of X & Y is given and I am not able to understand that whether it is top left, right or bottom left, right coordinate. I hope my question is clear and making sense to you guys. Please solve my dilemma. Thanks in advance.

", "votes": 0, "replies": [] } ] }, { "id": 376611, "author_name": "Kobe Python", "author_url": "", "post_date": "2018-08-27T19:41:31.207000", "content": "

Thanks Phil. I am unable to download the data. I get a network error if I try downloading it directly. Or a \"TqdmKeyError(\"Unknown argument(s): \" + str(kwargs)))\" error if I use the kaggle api. Is there a problem at your end or am I downloading it incorrectly.

", "votes": 0, "replies": [ { "id": 376616, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-08-27T19:50:39.067000", "content": "

Thanks. We'll check it out.

", "votes": 0, "replies": [] }, { "id": 376618, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-08-27T19:55:11", "content": "

Which version of the API are you using?

", "votes": 0, "replies": [] }, { "id": 376619, "author_name": "Kobe Python", "author_url": "", "post_date": "2018-08-27T19:58:38.763000", "content": "

Sorry for the trouble. Was able to download it on the 4th try.

", "votes": 0, "replies": [] }, { "id": 376626, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-08-27T20:46:44.743000", "content": "

I'm glad you resolved the problem.

", "votes": 0, "replies": [] } ] }, { "id": 376568, "author_name": "Dan Chernoff", "author_url": "", "post_date": "2018-08-27T18:30:12.730000", "content": "

Thanks Phil. The link to the starter kernel is bad - 404 error.

", "votes": 0, "replies": [ { "id": 376576, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-08-27T18:39:52.643000", "content": "

Thanks Dan! Should be available now.

", "votes": 0, "replies": [] }, { "id": 379137, "author_name": "YaGana Sheriff-Hussaini", "author_url": "", "post_date": "2018-08-30T23:38:06.780000", "content": "

What is the link to the starter code?

", "votes": 0, "replies": [] }, { "id": 381491, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-09-04T17:39:55.960000", "content": "

The page containing the starter code is here. It is also linked in the original post at the top of this thread.

", "votes": 0, "replies": [] }, { "id": 381585, "author_name": "YaGana Sheriff-Hussaini", "author_url": "", "post_date": "2018-09-04T20:11:25.833000", "content": "

Thanks.

", "votes": 0, "replies": [] } ] }, { "id": 399807, "author_name": "", "author_url": "", "post_date": "2018-10-06T20:13:24.633000", "content": "", "votes": 0, "replies": [] }, { "id": 379208, "author_name": "", "author_url": "", "post_date": "2018-08-31T02:45:07.080000", "content": "", "votes": 0, "replies": [ { "id": 379211, "author_name": "Phil Culliton", "author_url": "", "post_date": "2018-08-31T02:50:33.140000", "content": "

Some images have multiple bounding boxes. Each bounding box is a row. There are the same number of unique patientIds in the training set as there are images.

", "votes": 1, "replies": [] } ] }, { "id": 377069, "author_name": "", "author_url": "", "post_date": "2018-08-28T14:45:03.960000", "content": "", "votes": 0, "replies": [] }, { "id": 377174, "author_name": "VijayaKakumani", "author_url": "", "post_date": "2018-08-28T17:23:01.297000", "content": "

Thank you Phil.

", "votes": 0, "replies": [] }, { "id": 376823, "author_name": "Harini Perera", "author_url": "", "post_date": "2018-08-28T05:49:54.280000", "content": "

Thanks Phil.

", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "376495": "Welcome to RSNA Pneumonia Detection Challenge!\n\nIn this competition, we are finding and localizing pneumonia in images of human lungs. You should familiarize yourself with the competition metric - as well as the concept of Intersection over Union if you haven't encountered it - and the file descriptions on the [Data][1] tab.\n\nAlso be sure to check out our excellent Getting Started resources at:\n\nhttps://www.kaggle.com/c/rsna-pneumonia-detection-challenge#Getting%20Started\n\nWe have a kernel for exploratory data analysis, as well as a starter notebook that walks through getting the data loaded and modeled. Both provide insights on the data that competitors won't want to miss.\n\nNote that images can contain **multiple bounding boxes**. Each box is represented by a row in the training data, but (as described in the [Evaluation][2] page) your submission should have one row per image that includes all of its bounding boxes. Please check the [Evaluation][3] and [Data][4] pages for a more detailed description of how this works.\n\nWe'll start a separate thread in a few days for domain-specific questions. In the meantime, please feel free to ask your questions in this thread.\n\nBest of luck to you all!\n\n\n [1]: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/data\n [2]: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge#evaluation\n [3]: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge#evaluation\n [4]: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/data", "383294": "although the information in the evaluation page is probably enough it would be fantastic if the host or support team will release a github or kernel with official evaluation matrix that we can run against subset of gt. I would guess that in the long run it would save us millions of submissions and millions of questions for the host team.\nA python script that takes gt format csv and predictions csv and calculate score would really really help.", "380154": "The evaluation script is really needed - https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64542", "378690": "Hi @philculliton, if an image has no ground-truth bounding box, then no matter what the prediction string is (even leaving as blank), the score for that image is always 0, is that right? ", "377439": "In the light of recent leak from \"Airbus Ship Detection Challenge\" and based on the fact that in the description is stipulated that the data source is \"[NIHCC][1]\" that contains more labeled images than the current set - Can the organizers confirm that the \"test\" data is not based on \"train\" data in NIHCC?\n\n\n [1]: https://nihcc.app.box.com/v/ChestXray-NIHCC/folder/36938765345", "376739": "Hi Phil, I have two questions: \ncould you explain the \"confidence\" number that should be submitted with each bounding box. should it be between 0 and 1?\nAlso, in the rules I see that the \"model\" is required to be submitted by the end of stage 1. Is \"model\" means the source code? Does it matter what programming software we are using?\n\nthanks.\n ", "391109": "can you please address https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/66323 ?", "1121534": "What type of preprocessing should I do for those X-ray Images and what is best technique to find the opacity ? \nPlease anyone help me regarding this :)", "755151": "Great dataset! I'm very interested in this competition!", "525984": "How can i download dataset?\n", "405712": "our team missed the team merging deadline... we were too busy with developing the model. Is there any way for us to ask for a team merge privately? Hope to get a reply. Thank you", "405342": "I have a question about the evaluation metric. If we predict a box with IoU below the threshold, would the box be counted as a false negative or a false positive? What if we also predicted another box that had better IoU with the same ground truth box? Would it then be counted as best of the 2 or as 1 true positive and 1 false positive?", "404355": "Great dataset! An awesome competition for those who want to start applying DL for medical data. ", "390999": "writing it just to lift my rank....sorry", "381648": "Great dataset! I'm looking forward to joining the competition!", "380626": "I know, i shouldn't ask this , but is it necessary to download this 3gb of data or any other option is available .", "380142": "Could you please clarify on this - https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64644\n?", "379210": "Hi Phil,\nWould you please confirm the number of training images in stage_1_train_images.zip? After unzipping, I'm showing there are 25,685 dcm files while the stage_1_train_labels.csv file has 28,989 patient entries. Thanks for a great contest!", "379051": "I'm wondering about the GCP credits. Under \"Getting Started\" it says \"View the Data page for the URL\" but I see no URL there. Is this because:\n\n 1. All credits have already been awarded, and the URL has been removed;\n 2. The URL hasn't been added yet; \n 3. It is only visible to people who have already made a submission; or\n 4. I didn't look closely enough?\n\n(Note: I have accepted the rules but have not made a submission. Have not decided whether to do so yet, but the \"get 'em before they run out\" offer would certainly be an issue in that decision.)", "378682": "Hi Phil. Should I upload (and license OpenSource if I win) my training code, as well as inference code? Or, if my uploaded model and inference code can reproduce the result, do I only have to upload them? Thank you!", "376714": "Hi, Phill!\nMay I confirm about the label data? (stage_1_train_labels.csv)\n\nI found some duplicated patient ID in labels. \nThat is what you mentioned about the multiple bouncing box, doesn't it? And Does it just mean the label is stored in multiple rows in train_labels, doesn't it?\n(For example, patientId = '7d674c82-5501-4730-92c5-d241fd6911e7' appears 4 times in labels.csv file.)\n\nThanks.", "376611": "Thanks Phil. I am unable to download the data. I get a network error if I try downloading it directly. Or a \"TqdmKeyError(\"Unknown argument(s): \" + str(kwargs)))\" error if I use the kaggle api. Is there a problem at your end or am I downloading it incorrectly.", "376568": "Thanks Phil. The link to the starter kernel is bad - 404 error.", "399807": "", "379208": "", "377069": "", "377174": "Thank you Phil.", "376823": "Thanks Phil." } }