Per the Competition Rules, publicly and freely available external data is permitted, if it is available for use that includes research or academic purposes. If you are a winner, your solution must also be released under the stated open source license, per the rules' winner license requirements.
\nAll external data must be posted to this forum thread no later than the Entry Deadline (one week before competition close).
\nOnce someone posts an external dataset to this thread, you do not need to re-post it if you are using the same one.
\nYou only need to declare the original dataset used; you do not need to declare re-labeled or augmented or otherwise processed versions of datasets. Pre-trained models can be declared, as well; however, models resulting from your own original work, that you have trained yourself offline do not need to be shared/declared.
", "messages": [ { "id": "919240", "postDate": "07/07/2020 19:09:50", "content": "Per the Competition Rules, publicly and freely available external data is permitted, if it is available for use that includes research or academic purposes. If you are a winner, your solution must also be released under the stated open source license, per the rules' winner license requirements.
\nAll external data must be posted to this forum thread no later than the Entry Deadline (one week before competition close).
\nOnce someone posts an external dataset to this thread, you do not need to re-post it if you are using the same one.
\nYou only need to declare the original dataset used; you do not need to declare re-labeled or augmented or otherwise processed versions of datasets. Pre-trained models can be declared, as well; however, models resulting from your own original work, that you have trained yourself offline do not need to be shared/declared.
", "rawMarkdown": "Per the [Competition Rules](https://www.kaggle.com/c/osic-pulmonary-fibrosis-progression/rules), **publicly and freely available external data is permitted, if it is available for use that includes research or academic purposes**. If you are a winner, your solution must also be released under the stated open source license, per the rules' winner license requirements. \n\nAll external data must be posted to this forum thread no later than the Entry Deadline (one week before competition close).\n\nOnce someone posts an external dataset to this thread, you do not need to re-post it if you are using the same one.\n\nYou only need to declare the original dataset used; you do not need to declare re-labeled or augmented or otherwise processed versions of datasets. Pre-trained models can be declared, as well; however, models resulting from your own original work, that you have trained yourself offline do not need to be shared/declared.", "votes": null }, { "id": "921754", "postDate": "07/09/2020 14:16:16", "content": "Can models for pre-processing (e.g. segmentation) be used if they were trained on private data? Is it required that such a model is declared and made publicly available prior to the entry deadline?
", "rawMarkdown": "Can models for pre-processing (e.g. segmentation) be used if they were trained on private data? Is it required that such a model is declared and made publicly available prior to the entry deadline?", "votes": null }, { "id": "921905", "postDate": "07/09/2020 16:36:03", "content": "@hojijoji If the pre-trained model itself is freely and publicly licensed for \"use that includes research and academic purposes,\" and the resulting solution/code could be open source licensed (including for commercial use) if it in a winning position, then yes, it can be used. However, if licensing prevents these requirements from being met, then it cannot be used. And yes, the pre-trained model should be declared prior to the entry deadline.
", "rawMarkdown": "hojijoji If the pre-trained model itself is freely and publicly licensed for \"use that includes research and academic purposes,\" and the resulting solution/code could be open source licensed (including for commercial use) if it in a winning position, then yes, it can be used. However, if licensing prevents these requirements from being met, then it cannot be used. And yes, the pre-trained model should be declared prior to the entry deadline.", "votes": null }, { "id": "925231", "postDate": "07/11/2020 23:56:56", "content": "The Dataset http://medgift.hevs.ch/wordpress/databases/ild-database/, it has the the following description:\n> The database is available for research on request and after signature of a license agreement. \nyou have to sign and send the licensed agreement by email then you can access the data.
\n\nAlthough it's not readily available, it is so by request for research and academic purposes.
\n\n@juliaelliott Can you confirm it if it abides by the competition rules?
\n\nThank you in advance
", "rawMarkdown": "The Dataset http://medgift.hevs.ch/wordpress/databases/ild-database/, it has the the following description:\n> The database is available for research on request and after signature of a license agreement. \nyou have to sign and send the licensed agreement by email then you can access the data.\n\nAlthough it's not readily available, it is so by request for research and academic purposes.\n\n@juliaelliott Can you confirm it if it abides by the competition rules?\n\nThank you in advance", "votes": null }, { "id": "930721", "postDate": "07/15/2020 17:07:26", "content": "Well, by consulting the licensed agreement we would have to sign in the website above, it says:\n> The End-User named must be a legal institution, or a department or section of a named legal institution, not an individual nor a project -the person signing this Agreement has to be duly authorized by the institution for such signatures (e.g., Department or Administrative Head or similar) and shall be liable for such authorization.
\n\nSince we Kagglers would be classified as persons and not a Research Institution, I didn't bother to request access to the Database mentioned and abandoned the effort. So the Dataset will not be used for me and the message can be ignored.
\n\nIf anyone else succeeds, please let me know.
", "rawMarkdown": "Well, by consulting the licensed agreement we would have to sign in the website above, it says:\n> The End-User named must be a legal institution, or a department or section of a named legal institution, not an individual nor a project -the person signing this Agreement has to be duly authorized by the institution for such signatures (e.g., Department or Administrative Head or similar) and shall be liable for such authorization.\n\nSince we Kagglers would be classified as persons and not a Research Institution, I didn't bother to request access to the Database mentioned and abandoned the effort. So the Dataset will **not** be used for me and the message can be ignored.\n\nIf anyone else succeeds, please let me know.", "votes": null }, { "id": "930743", "postDate": "07/15/2020 17:21:05", "content": "@ronaldokun Based on your observation regarding a gating mechanism on the dataset which would make its access and availability not fully public to all other competitors, this would be prohibited by the rules. If, however, there are research-use-licensed datasets which are not gated (and therefore freely/publicly available to all competitors), the host has specified that such datasets are permitted, so long as the resulting solution/code being created, if in winning standing, will be licensed as open source, including commercial use.
", "rawMarkdown": "ronaldokun Based on your observation regarding a gating mechanism on the dataset which would make its access and availability not fully public to all other competitors, this would be prohibited by the rules. If, however, there are research-use-licensed datasets which are not gated (and therefore freely/publicly available to all competitors), the host has specified that such datasets are permitted, so long as the resulting solution/code being created, if in winning standing, will be licensed as open source, including commercial use.", "votes": null }, { "id": "930745", "postDate": "07/15/2020 17:22:26", "content": "Thank you Julia for the clarification.
", "rawMarkdown": "Thank you Julia for the clarification.", "votes": null }, { "id": "944273", "postDate": "07/25/2020 01:47:19", "content": "Computed Tomography Emphysema Database
\n\nLink: https://lauge-soerensen.github.io/emphysema-database/
\n\nAs per authors is free of charge for research and educational purposes. Uses TIFF.
", "rawMarkdown": "**Computed Tomography Emphysema Database**\n\nLink: https://lauge-soerensen.github.io/emphysema-database/\n\nAs per authors is free of charge for research and educational purposes. Uses TIFF.", "votes": null }, { "id": "944289", "postDate": "07/25/2020 02:02:09", "content": "Emphysema is a different disease
", "rawMarkdown": "Emphysema is a different disease", "votes": null }, { "id": "944359", "postDate": "07/25/2020 03:44:40", "content": "That is true. But there are overlaps with the two diseases, including an entity called \"Combined Pulmonary Fibrosis and Emphysema Syndrome.\"
\n\n\n\n\nRadiographic characteristics include the presence of upper-lobe emphysema and lower-lobe pulmonary fibrosis.
\n
This paper also notes that the proportion of patients with combined PF and emphysema can range from 18.8% to 50.9%. It seems probable that a proportion of the patients will possibly have this component as well. Note that these patients are said to experience slower decline in FVC (which elucidating is a clear goal of this competition). Either way, it seems having the ability to identify radiologic findings consistent with emphysema and then have that as a feature may be helpful given what the literature states. YMMV
", "rawMarkdown": "That is true. But there are overlaps with the two diseases, including an entity called \"[Combined Pulmonary Fibrosis and Emphysema Syndrome](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3251269/).\" \n\n> Radiographic characteristics include the presence of upper-lobe emphysema and lower-lobe pulmonary fibrosis.\n\nThis paper also notes that the proportion of patients with combined PF and emphysema can range from 18.8% to 50.9%. It seems probable that a proportion of the patients will possibly have this component as well. Note that these patients are said to experience slower decline in FVC (which elucidating is a clear goal of this competition). Either way, it seems having the ability to identify radiologic findings consistent with emphysema and then have that as a feature may be helpful given what the literature states. YMMV", "votes": null }, { "id": "944693", "postDate": "07/25/2020 09:02:14", "content": "I am using lungmask, licensed under the GNU Public License
\n\nJohannes Hofmanninger, Forian Prayer, Jeanny Pan, Sebastian Röhrich, Helmut Prosch and Georg Langs. \"Automatic lung segmentation in routine imaging is a data diversity problem, not a methodology problem\". 1 2020, https://arxiv.org/abs/2001.11767
", "rawMarkdown": "I am using lungmask, licensed under the GNU Public License\n\nJohannes Hofmanninger, Forian Prayer, Jeanny Pan, Sebastian Röhrich, Helmut Prosch and Georg Langs. \"Automatic lung segmentation in routine imaging is a data diversity problem, not a methodology problem\". 1 2020, https://arxiv.org/abs/2001.11767", "votes": null }, { "id": "955933", "postDate": "08/03/2020 05:15:50", "content": "About (2.) - I want to create this offline, as it takes long time, upload it somewhere to make it public and share it with others so it will be public and usable
", "rawMarkdown": "1. If kernels have no internet access, how am I to use external data whether of third party or my own?\n2. If I create my own dataset (segmented version of the images) and I want to share it publicly (so we all can use it) - what is the procedure to do so?\n\nAbout (2.) - I want to create this offline, as it takes long time, upload it somewhere to make it public and share it with others so it will be public and usable", "votes": null }, { "id": "955940", "postDate": "08/03/2020 05:20:56", "content": "Using the same, to be specific to an implementation (as it is considered data) - https://github.com/JoHof/lungmask
", "rawMarkdown": "Using the same, to be specific to an implementation (as it is considered data) - https://github.com/JoHof/lungmask", "votes": null }, { "id": "957705", "postDate": "08/04/2020 13:58:13", "content": "Have you or anyone run this on Kaggle without internet? do respond
", "rawMarkdown": "Have you or anyone run this on Kaggle without internet? do respond", "votes": null }, { "id": "966270", "postDate": "08/11/2020 09:47:05", "content": "I was wondering: are we allowed to train our models on our own machines, and then simply let the notebook do the predictions?
\n\nLet's say I wanted to use an auto-encoder to handle the images, and then use another prediction model to combine the data from the images, with the FVC data, could I train the autoencoder on my own machine, then add the trained model as a resource and load it in my notebook to do the rest of the pipeline?\nOr does any training with the actual data have to happen in the notebook within those time constraints?
\n\nThank you
", "rawMarkdown": "I was wondering: are we allowed to train our models on our own machines, and then simply let the notebook do the predictions?\n\nLet's say I wanted to use an auto-encoder to handle the images, and then use another prediction model to combine the data from the images, with the FVC data, could I train the autoencoder on my own machine, then add the trained model as a resource and load it in my notebook to do the rest of the pipeline?\nOr does any training with the actual data have to happen in the notebook within those time constraints?\n\nThank you", "votes": null }, { "id": "966862", "postDate": "08/11/2020 17:41:47", "content": "@stadlerm Yes, you can train locally and upload the trained model into Kaggle's notebook as an external data source, effectively requiring that you minimally run inference in your notebook submission on Kaggle. The code must be possible to re-run on a hidden test set swapped out for the current test set.
", "rawMarkdown": "stadlerm Yes, you can train locally and upload the trained model into Kaggle's notebook as an external data source, effectively requiring that you minimally run inference in your notebook submission on Kaggle. The code must be possible to re-run on a hidden test set swapped out for the current test set.", "votes": null }, { "id": "967776", "postDate": "08/12/2020 13:55:32", "content": "Is open source licensing okay for you?
", "rawMarkdown": "Is open source licensing okay for you?", "votes": null }, { "id": "968035", "postDate": "08/12/2020 16:42:32", "content": "@kurianbenoy - Yes, if external data is open source licensed and freely & publicly available (no cost to any other competitors), then it is permitted. You would need to declare the external data you are using on this forum thread before the entry deadline.
", "rawMarkdown": "kurianbenoy - Yes, if external data is open source licensed and freely & publicly available (no cost to any other competitors), then it is permitted. You would need to declare the external data you are using on this forum thread before the entry deadline.", "votes": null }, { "id": "991757", "postDate": "08/30/2020 16:35:38", "content": "I did run this without the internet, by creating a dataset. I compared it with other segmentation methods in this kernel
\nhttps://www.kaggle.com/gautham11/lung-segmentation-methods-comparison
\nBut I can't use this in any submission, because it takes ~2 seconds to segment an image. Let me know if anyone managed to get around this.
\nWatershed segmentation seems like the way to go.
Are these licenses allowed in this competition? (CC BY NC 3.0, CC BY NC SA 3.0, CC BY 4.0)
", "rawMarkdown": "Are these licenses allowed in this competition? **(CC BY NC 3.0, CC BY NC SA 3.0, CC BY 4.0)**", "votes": null }, { "id": "995586", "postDate": "09/02/2020 15:35:18", "content": "@orkatz2 The competition permits publicly and freely available external data is permitted if it is available for use that includes research or academic purposes. These all appear to be licenses that are available for research/academic purposes, however you should also ensure they are free to all and publicly available.
", "rawMarkdown": "orkatz2 The competition permits publicly and freely available external data is permitted if it is available for use that includes research or academic purposes. These all appear to be licenses that are available for research/academic purposes, however you should also ensure they are free to all and publicly available.", "votes": null }, { "id": "997419", "postDate": "09/04/2020 03:27:49", "content": "A model related to segmentation
\nHere is an interesting link to the segmentation of biological images including Lung using the BCDU network.
\nhttps://github.com/rezazad68/BCDU-Net
\nIt seems it is freely available according to the author, He only mentioned CITE them. I haven't found any Licence file in the GitHub. May I use them?
@nishyanand If you cannot properly identify provenance of and licensing of the dataset, I would recommend not using it.
", "rawMarkdown": "nishyanand If you cannot properly identify provenance of and licensing of the dataset, I would recommend not using it.", "votes": null }, { "id": "998651", "postDate": "09/04/2020 22:50:30", "content": "I have sent an email to the author(rezazad68@gmail.com is the email address of the first author mentioned in the GitHub link) and got a response as
\n\"Code and weight are open access just cite the paper in case of using them.\".
\n I have attached the email conversation as well. I hope this is enough for use it. @juliaelliott please let me know.
@nishyanand This sounds acceptable.
", "rawMarkdown": "nishyanand This sounds acceptable.", "votes": null }, { "id": "1009306", "postDate": "09/13/2020 20:48:00", "content": "I have found that there are several R packages that are in CRAN but for some reason are missing from the notebook R script environment. But I am able to upload them via \"Add data\" as a zip file and then execute the commands in my script to unpack and install them. Do I need to declare these \"datasets\" (zip files of R packages) as external data in this thread?
", "rawMarkdown": "I have found that there are several R packages that are in CRAN but for some reason are missing from the notebook R script environment. But I am able to upload them via \"Add data\" as a zip file and then execute the commands in my script to unpack and install them. Do I need to declare these \"datasets\" (zip files of R packages) as external data in this thread?", "votes": null }, { "id": "1010483", "postDate": "09/14/2020 19:27:31", "content": "lungmask https://github.com/JoHof/lungmask
", "rawMarkdown": "lungmask https://github.com/JoHof/lungmask", "votes": null }, { "id": "1011058", "postDate": "09/15/2020 08:17:14", "content": "if I train locally and then upload the weights, do I have to share my weights here as external data source or make it open source?
", "rawMarkdown": "if I train locally and then upload the weights, do I have to share my weights here as external data source or make it open source?", "votes": null }, { "id": "1015668", "postDate": "09/18/2020 10:46:30", "content": "Can i use the DataSet learning from this contest for a Research paper by citing the Data Source?
", "rawMarkdown": "Can i use the DataSet learning from this contest for a Research paper by citing the Data Source?", "votes": null }, { "id": "1015998", "postDate": "09/18/2020 15:35:24", "content": "@siddarthpaim Yes, you may use the dataset in compliance with the data use terms (Section 7) in the competition rules.
", "rawMarkdown": "siddarthpaim Yes, you may use the dataset in compliance with the [data use terms (Section 7) in the competition rules](https://www.kaggle.com/c/osic-pulmonary-fibrosis-progression/rules).", "votes": null }, { "id": "1016003", "postDate": "09/18/2020 15:36:25", "content": "@dslate Packages are not necessary to declare. Pre-trained models are.
", "rawMarkdown": "dslate Packages are not necessary to declare. Pre-trained models are.", "votes": null }, { "id": "1027533", "postDate": "09/26/2020 08:15:42", "content": "(https://github.com/JoHof/lungmask)
\nAutomated lung segmentation in CT
Pretrained models from:
\nhttps://pytorch.org/docs/stable/torchvision/models.html
\nhttps://github.com/rwightman/pytorch-image-models/
I am using below pretrained model
\nhttps://www.kaggle.com/khoongweihao/osic-model-weights
Hi @juliaelliott , is GPLv3 allowed? I read on thread here, which looks like it is, commercial and privates are both permitted, but I didn't see this license being mentioned anywhere else in the forum, so just want to confirm it. Thanks.
", "rawMarkdown": "Hi @juliaelliott , is GPLv3 allowed? I read on thread [here](https://www.kaggle.com/c/osic-pulmonary-fibrosis-progression/discussion/180026), which looks like it is, commercial and privates are both permitted, but I didn't see this license being mentioned anywhere else in the forum, so just want to confirm it. Thanks.", "votes": null }, { "id": "1030469", "postDate": "09/28/2020 17:23:59", "content": "@samshipengs
\nIs it publicly available at no cost to other participants?
\nIs it licensed for use that includes academic/research purposes?
\nIf you use the dataset, is it possible for you to license your solution/code in accordance with the competition's \"Winners License\" rules?
If the answer to those 3 questions is yes, then yes, it is permitted. If any answer is no, then no, it is not permitted.
", "rawMarkdown": "samshipengs \nIs it publicly available at no cost to other participants?\nIs it licensed for use that includes academic/research purposes?\nIf you use the dataset, is it possible for you to license your solution/code in accordance with the competition's \"Winners License\" rules?\n\nIf the answer to those 3 questions is yes, then yes, it is permitted. If any answer is no, then no, it is not permitted.", "votes": null }, { "id": "1030598", "postDate": "09/28/2020 18:56:30", "content": "@samshipengs actually I looked at the link in that thread in more detail and found something. You can see my response here.
\n@juliaelliott could you please have a quick look also? I understand that your three question system is meant to make things simple, but that third question:
\n\n\nIf you use the dataset, is it possible for you to license your solution/code in accordance with the competition's \"Winners License\" rules?
\n
is actually really involved in terms of understanding copyright laws. Wondering if there is anyone on your team who can definitively tell us whether we may use and/or adapt GPLv3 licensed code.
", "rawMarkdown": "samshipengs actually I looked at the link in that thread in more detail and found something. You can see my response [here](https://www.kaggle.com/c/osic-pulmonary-fibrosis-progression/discussion/180026#1030594).\n\n@juliaelliott could you please have a quick look also? I understand that your three question system is meant to make things simple, but that third question:\n\n> If you use the dataset, is it possible for you to license your solution/code in accordance with the competition's \"Winners License\" rules?\n\nis actually really involved in terms of understanding copyright laws. Wondering if there is anyone on your team who can definitively tell us whether we may use and/or adapt GPLv3 licensed code.", "votes": null }, { "id": "1030614", "postDate": "09/28/2020 19:12:55", "content": "https://www.kaggle.com/andrewmvd/covid19-ct-scans
", "rawMarkdown": "https://www.kaggle.com/andrewmvd/covid19-ct-scans", "votes": null }, { "id": "1030660", "postDate": "09/28/2020 20:18:25", "content": "@alexandersoare Thanks for your question in that thread, actually that part was exactly what I didnt understand when I was reading the link shared by @stanleyjzheng , so I though its better to check with officials. just for reference, this is the part i didnt understand:
\n\n\nNote that compatibility with the GPLv3 is one-way only, which means you may license your contributions to adaptations of BY-SA 4.0 materials under GPLv3, but you may not license your contributions to adaptations of GPLv3 projects under BY-SA 4.0. Other special considerations apply. See the full analysis and comparison for more information.
\n
My attempt at this is, it means you can use GPLv3 code for BY-SA4 projects but not the other way around (i.e. you can't use BY-SA4 code for GPLv3 projects), is that correct?
\n@juliaelliott thanks for your guide on checking allowed datasets, I think the answer to the first two questions are affirmative, its no cost for everyone to use and i suppose there is no issue to use for academic/research since its even okay for commercial use, but im really not sure for the third question. Would really appreciate if you could provide more clarity.
", "rawMarkdown": "alexandersoare Thanks for your question in that thread, actually that part was exactly what I didnt understand when I was reading the link shared by @stanleyjzheng , so I though its better to check with officials. just for reference, this is the part i didnt understand:\n> Note that compatibility with the GPLv3 is one-way only, which means you may license your contributions to adaptations of BY-SA 4.0 materials under GPLv3, but you may not license your contributions to adaptations of GPLv3 projects under BY-SA 4.0. Other special considerations apply. See the full analysis and comparison for more information. \n\nMy attempt at this is, it means you can use GPLv3 code for BY-SA4 projects but not the other way around (i.e. you can't use BY-SA4 code for GPLv3 projects), is that correct? \n\n@juliaelliott thanks for your guide on checking allowed datasets, I think the answer to the first two questions are affirmative, its no cost for everyone to use and i suppose there is no issue to use for academic/research since its even okay for commercial use, but im really not sure for the third question. Would really appreciate if you could provide more clarity.", "votes": null }, { "id": "1030692", "postDate": "09/28/2020 21:02:41", "content": "@samshipengs If using that dataset prohibits you to license your final solution as required by the competition rules, then the 3rd question I posed would be no, and that would make it prohibited for use.
", "rawMarkdown": "samshipengs If using that dataset prohibits you to license your final solution as required by the competition rules, then the 3rd question I posed would be no, and that would make it prohibited for use.", "votes": null }, { "id": "1030695", "postDate": "09/28/2020 21:18:26", "content": "Thanks for the reply, although the issue is I'm not sure if that data set would prohibit me to license my solution as required by the rules (I referenced above) and hoping to get clarification. Is it just me not understanding it? if so could someone help to explain?
\n@hojijoji Would you share some insights if you don't mind? (I believe you are the author of the repo, thanks for the great work!)
", "rawMarkdown": "Thanks for the reply, although the issue is I'm not sure if that data set would prohibit me to license my solution as required by the rules (I referenced above) and hoping to get clarification. Is it just me not understanding it? if so could someone help to explain?\n\n@hojijoji Would you share some insights if you don't mind? (I believe you are the author of the repo, thanks for the great work!)", "votes": null }, { "id": "1030731", "postDate": "09/28/2020 22:45:17", "content": "@juliaelliott may we use the scans currently available in the RSNA STR Pulmonary Embolism Detection\n competition?
\nParticularly for unsupervised learning
", "rawMarkdown": "juliaelliott may we use the scans currently available in the [RSNA STR Pulmonary Embolism Detection\n](https://www.kaggle.com/c/rsna-str-pulmonary-embolism-detection) competition? \n\nParticularly for unsupervised learning", "votes": null }, { "id": "1030743", "postDate": "09/28/2020 23:05:23", "content": "https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf2_detection_zoo.md
\nhttps://github.com/UCSD-AI4H/COVID-CT
\nhttps://github.com/JoHof/lungmask
\nhttp://www.via.cornell.edu/crpf.html
\nhttps://www.kaggle.com/andrewmvd/covid19-ct-scans
\nhttps://wiki.cancerimagingarchive.net/display/Public/NSCLC+Radiogenomics
\nhttps://wiki.cancerimagingarchive.net/display/Public/RIDER+Lung+CT
\nhttps://wiki.cancerimagingarchive.net/display/Public/LIDC-IDRI
\nhttps://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=70224216
\nhttps://wiki.cancerimagingarchive.net/display/Public/Lung+Fused-CT-Pathology
\nhttps://wiki.cancerimagingarchive.net/display/Public/NSCLC-Radiomics
\nhttps://wiki.cancerimagingarchive.net/display/Public/NSCLC-Radiomics-Interobserver1
\nhttps://wiki.cancerimagingarchive.net/display/Public/TCGA-LUSC
\nhttps://wiki.cancerimagingarchive.net/display/Public/LungCT-Diagnosis
\nhttps://wiki.cancerimagingarchive.net/display/Public/NSCLC-Radiomics-Genomics
\nhttps://wiki.cancerimagingarchive.net/display/Public/SPIE-AAPM+Lung+CT+Challenge
\nhttps://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=52758026
\nhttps://wiki.cancerimagingarchive.net/display/Public/LIDC-IDRI
https://github.com/shijianjian/EfficientNet-PyTorch-3D
", "rawMarkdown": "https://github.com/shijianjian/EfficientNet-PyTorch-3D", "votes": null }, { "id": "1030750", "postDate": "09/28/2020 23:25:26", "content": "@tannercarbonati That competition's dataset is licensed for non-commercial purposes, so it should be eligible to use as long as you are not redistributing it.
", "rawMarkdown": "tannercarbonati That competition's dataset is licensed for non-commercial purposes, so it should be eligible to use as long as you are not redistributing it.", "votes": null }, { "id": "1031117", "postDate": "09/29/2020 08:53:23", "content": "@juliaelliott just to clarify this is not about datasets (from my end at least). It's about public code with a GPLv3 license. Surely someone will need to check when it comes time to evaluate the final solutions, so any chance we can get a clear answer now?
", "rawMarkdown": "juliaelliott just to clarify this is not about datasets (from my end at least). It's about public code with a GPLv3 license. Surely someone will need to check when it comes time to evaluate the final solutions, so any chance we can get a clear answer now?", "votes": null }, { "id": "1031125", "postDate": "09/29/2020 09:01:17", "content": "My understanding is that you can not relicense GPLv3 (copyleft) material under CC-BY 4.0 (permissive). Also, you should not confuse CC-BY (permissive) with CC-BY-SA (copyleft). However, eventhough I would have prefered to stick to a copyleft license, I have changed the repository to Apache 2.0 which should be ok with CC-BY 4.0 so that you can use it for this competition. Note, that this only affects to current version!
", "rawMarkdown": "My understanding is that you can not relicense GPLv3 (copyleft) material under CC-BY 4.0 (permissive). Also, you should not confuse CC-BY (permissive) with CC-BY-SA (copyleft). However, eventhough I would have prefered to stick to a copyleft license, I have changed the [repository ](https://github.com/JoHof/lungmask) to Apache 2.0 which should be ok with CC-BY 4.0 so that you can use it for this competition. Note, that this only affects to current version!", "votes": null }, { "id": "1031190", "postDate": "09/29/2020 09:35:47", "content": "@hojijoji many thanks!
", "rawMarkdown": "hojijoji many thanks!", "votes": null }, { "id": "1031301", "postDate": "09/29/2020 11:21:02", "content": "thanks for the explanation and modifying license for current version!
", "rawMarkdown": "thanks for the explanation and modifying license for current version!", "votes": null }, { "id": "1031422", "postDate": "09/29/2020 12:47:28", "content": "EfficientNet B0 to B7
\nhttps://github.com/qubvel/efficientnet
https://github.com/qubvel/efficientnet
\nhttps://github.com/JoHof/lungmask/
\nhttps://github.com/dlpbc/keras-kinetics-i3d
https://www.kaggle.com/bjaeger/pretrained-weights
", "rawMarkdown": "https://www.kaggle.com/bjaeger/pretrained-weights", "votes": null }, { "id": "1031915", "postDate": "09/29/2020 19:18:56", "content": "https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet
", "rawMarkdown": "https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet", "votes": null }, { "id": "1032054", "postDate": "09/29/2020 23:00:16", "content": "Hi. I am using pretrained models from repo https://github.com/lfz/DSB2017 which has open source license that was mentioned in the dsb2017 competition rules; cite of paper arXiv:1711.08324 will be done.
", "rawMarkdown": "Hi. I am using pretrained models from repo https://github.com/lfz/DSB2017 which has open source license that was mentioned in the dsb2017 competition rules; cite of paper arXiv:1711.08324 will be done.", "votes": null }, { "id": "1035642", "postDate": "10/02/2020 23:01:49", "content": "https://github.com/facebook/prophet
", "rawMarkdown": "https://github.com/facebook/prophet", "votes": null } ], "comments": [ { "id": 995399, "author_name": "orkatz2", "author_url": "", "post_date": "09/02/2020 12:08:21", "content": "Are these licenses allowed in this competition? (CC BY NC 3.0, CC BY NC SA 3.0, CC BY 4.0)
", "votes": null, "replies": [ { "id": 995586, "author_name": "juliaelliott", "author_url": "", "post_date": "09/02/2020 15:35:18", "content": "@orkatz2 The competition permits publicly and freely available external data is permitted if it is available for use that includes research or academic purposes. These all appear to be licenses that are available for research/academic purposes, however you should also ensure they are free to all and publicly available.
", "votes": null, "replies": [] } ] }, { "id": 997419, "author_name": "nishyanand", "author_url": "", "post_date": "09/04/2020 03:27:49", "content": "A model related to segmentation
\nHere is an interesting link to the segmentation of biological images including Lung using the BCDU network.
\nhttps://github.com/rezazad68/BCDU-Net
\nIt seems it is freely available according to the author, He only mentioned CITE them. I haven't found any Licence file in the GitHub. May I use them?
@nishyanand If you cannot properly identify provenance of and licensing of the dataset, I would recommend not using it.
", "votes": null, "replies": [] }, { "id": 998651, "author_name": "nishyanand", "author_url": "", "post_date": "09/04/2020 22:50:30", "content": "I have sent an email to the author(rezazad68@gmail.com is the email address of the first author mentioned in the GitHub link) and got a response as
\n\"Code and weight are open access just cite the paper in case of using them.\".
\n I have attached the email conversation as well. I hope this is enough for use it. @juliaelliott please let me know.
@nishyanand This sounds acceptable.
", "votes": null, "replies": [] } ] }, { "id": 1009306, "author_name": "dslate", "author_url": "", "post_date": "09/13/2020 20:48:00", "content": "I have found that there are several R packages that are in CRAN but for some reason are missing from the notebook R script environment. But I am able to upload them via \"Add data\" as a zip file and then execute the commands in my script to unpack and install them. Do I need to declare these \"datasets\" (zip files of R packages) as external data in this thread?
", "votes": null, "replies": [ { "id": 1016003, "author_name": "juliaelliott", "author_url": "", "post_date": "09/18/2020 15:36:25", "content": "@dslate Packages are not necessary to declare. Pre-trained models are.
", "votes": null, "replies": [] } ] }, { "id": 1010483, "author_name": "samshipengs", "author_url": "", "post_date": "09/14/2020 19:27:31", "content": "lungmask https://github.com/JoHof/lungmask
", "votes": null, "replies": [] }, { "id": 1015668, "author_name": "siddarthpaim", "author_url": "", "post_date": "09/18/2020 10:46:30", "content": "Can i use the DataSet learning from this contest for a Research paper by citing the Data Source?
", "votes": null, "replies": [ { "id": 1015998, "author_name": "juliaelliott", "author_url": "", "post_date": "09/18/2020 15:35:24", "content": "@siddarthpaim Yes, you may use the dataset in compliance with the data use terms (Section 7) in the competition rules.
", "votes": null, "replies": [] } ] }, { "id": 1027533, "author_name": "riyasvk", "author_url": "", "post_date": "09/26/2020 08:15:42", "content": "(https://github.com/JoHof/lungmask)
\nAutomated lung segmentation in CT
Pretrained models from:
\nhttps://pytorch.org/docs/stable/torchvision/models.html
\nhttps://github.com/rwightman/pytorch-image-models/
I am using below pretrained model
\nhttps://www.kaggle.com/khoongweihao/osic-model-weights
Hi @juliaelliott , is GPLv3 allowed? I read on thread here, which looks like it is, commercial and privates are both permitted, but I didn't see this license being mentioned anywhere else in the forum, so just want to confirm it. Thanks.
", "votes": null, "replies": [ { "id": 1030469, "author_name": "juliaelliott", "author_url": "", "post_date": "09/28/2020 17:23:59", "content": "@samshipengs
\nIs it publicly available at no cost to other participants?
\nIs it licensed for use that includes academic/research purposes?
\nIf you use the dataset, is it possible for you to license your solution/code in accordance with the competition's \"Winners License\" rules?
If the answer to those 3 questions is yes, then yes, it is permitted. If any answer is no, then no, it is not permitted.
", "votes": null, "replies": [] }, { "id": 1030598, "author_name": "alexandersoare", "author_url": "", "post_date": "09/28/2020 18:56:30", "content": "@samshipengs actually I looked at the link in that thread in more detail and found something. You can see my response here.
\n@juliaelliott could you please have a quick look also? I understand that your three question system is meant to make things simple, but that third question:
\n\n\nIf you use the dataset, is it possible for you to license your solution/code in accordance with the competition's \"Winners License\" rules?
\n
is actually really involved in terms of understanding copyright laws. Wondering if there is anyone on your team who can definitively tell us whether we may use and/or adapt GPLv3 licensed code.
", "votes": null, "replies": [] }, { "id": 1030660, "author_name": "samshipengs", "author_url": "", "post_date": "09/28/2020 20:18:25", "content": "@alexandersoare Thanks for your question in that thread, actually that part was exactly what I didnt understand when I was reading the link shared by @stanleyjzheng , so I though its better to check with officials. just for reference, this is the part i didnt understand:
\n\n\nNote that compatibility with the GPLv3 is one-way only, which means you may license your contributions to adaptations of BY-SA 4.0 materials under GPLv3, but you may not license your contributions to adaptations of GPLv3 projects under BY-SA 4.0. Other special considerations apply. See the full analysis and comparison for more information.
\n
My attempt at this is, it means you can use GPLv3 code for BY-SA4 projects but not the other way around (i.e. you can't use BY-SA4 code for GPLv3 projects), is that correct?
\n@juliaelliott thanks for your guide on checking allowed datasets, I think the answer to the first two questions are affirmative, its no cost for everyone to use and i suppose there is no issue to use for academic/research since its even okay for commercial use, but im really not sure for the third question. Would really appreciate if you could provide more clarity.
", "votes": null, "replies": [] }, { "id": 1030692, "author_name": "juliaelliott", "author_url": "", "post_date": "09/28/2020 21:02:41", "content": "@samshipengs If using that dataset prohibits you to license your final solution as required by the competition rules, then the 3rd question I posed would be no, and that would make it prohibited for use.
", "votes": null, "replies": [] }, { "id": 1030695, "author_name": "samshipengs", "author_url": "", "post_date": "09/28/2020 21:18:26", "content": "Thanks for the reply, although the issue is I'm not sure if that data set would prohibit me to license my solution as required by the rules (I referenced above) and hoping to get clarification. Is it just me not understanding it? if so could someone help to explain?
\n@hojijoji Would you share some insights if you don't mind? (I believe you are the author of the repo, thanks for the great work!)
", "votes": null, "replies": [] }, { "id": 1031117, "author_name": "alexandersoare", "author_url": "", "post_date": "09/29/2020 08:53:23", "content": "@juliaelliott just to clarify this is not about datasets (from my end at least). It's about public code with a GPLv3 license. Surely someone will need to check when it comes time to evaluate the final solutions, so any chance we can get a clear answer now?
", "votes": null, "replies": [] }, { "id": 1031125, "author_name": "hojijoji", "author_url": "", "post_date": "09/29/2020 09:01:17", "content": "My understanding is that you can not relicense GPLv3 (copyleft) material under CC-BY 4.0 (permissive). Also, you should not confuse CC-BY (permissive) with CC-BY-SA (copyleft). However, eventhough I would have prefered to stick to a copyleft license, I have changed the repository to Apache 2.0 which should be ok with CC-BY 4.0 so that you can use it for this competition. Note, that this only affects to current version!
", "votes": null, "replies": [] }, { "id": 1031190, "author_name": "alexandersoare", "author_url": "", "post_date": "09/29/2020 09:35:47", "content": "@hojijoji many thanks!
", "votes": null, "replies": [] }, { "id": 1031301, "author_name": "samshipengs", "author_url": "", "post_date": "09/29/2020 11:21:02", "content": "thanks for the explanation and modifying license for current version!
", "votes": null, "replies": [] } ] }, { "id": 1030614, "author_name": "orkatz2", "author_url": "", "post_date": "09/28/2020 19:12:55", "content": "https://www.kaggle.com/andrewmvd/covid19-ct-scans
", "votes": null, "replies": [] }, { "id": 1030731, "author_name": "tannercarbonati", "author_url": "", "post_date": "09/28/2020 22:45:17", "content": "@juliaelliott may we use the scans currently available in the RSNA STR Pulmonary Embolism Detection\n competition?
\nParticularly for unsupervised learning
", "votes": null, "replies": [ { "id": 1030750, "author_name": "juliaelliott", "author_url": "", "post_date": "09/28/2020 23:25:26", "content": "@tannercarbonati That competition's dataset is licensed for non-commercial purposes, so it should be eligible to use as long as you are not redistributing it.
", "votes": null, "replies": [] } ] }, { "id": 1030743, "author_name": "adityachhabra", "author_url": "", "post_date": "09/28/2020 23:05:23", "content": "https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf2_detection_zoo.md
\nhttps://github.com/UCSD-AI4H/COVID-CT
\nhttps://github.com/JoHof/lungmask
\nhttp://www.via.cornell.edu/crpf.html
\nhttps://www.kaggle.com/andrewmvd/covid19-ct-scans
\nhttps://wiki.cancerimagingarchive.net/display/Public/NSCLC+Radiogenomics
\nhttps://wiki.cancerimagingarchive.net/display/Public/RIDER+Lung+CT
\nhttps://wiki.cancerimagingarchive.net/display/Public/LIDC-IDRI
\nhttps://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=70224216
\nhttps://wiki.cancerimagingarchive.net/display/Public/Lung+Fused-CT-Pathology
\nhttps://wiki.cancerimagingarchive.net/display/Public/NSCLC-Radiomics
\nhttps://wiki.cancerimagingarchive.net/display/Public/NSCLC-Radiomics-Interobserver1
\nhttps://wiki.cancerimagingarchive.net/display/Public/TCGA-LUSC
\nhttps://wiki.cancerimagingarchive.net/display/Public/LungCT-Diagnosis
\nhttps://wiki.cancerimagingarchive.net/display/Public/NSCLC-Radiomics-Genomics
\nhttps://wiki.cancerimagingarchive.net/display/Public/SPIE-AAPM+Lung+CT+Challenge
\nhttps://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=52758026
\nhttps://wiki.cancerimagingarchive.net/display/Public/LIDC-IDRI
https://github.com/shijianjian/EfficientNet-PyTorch-3D
", "votes": null, "replies": [] }, { "id": 1031422, "author_name": "ramanareddyeceb", "author_url": "", "post_date": "09/29/2020 12:47:28", "content": "EfficientNet B0 to B7
\nhttps://github.com/qubvel/efficientnet
https://github.com/qubvel/efficientnet
\nhttps://github.com/JoHof/lungmask/
\nhttps://github.com/dlpbc/keras-kinetics-i3d
https://www.kaggle.com/bjaeger/pretrained-weights
", "votes": null, "replies": [] }, { "id": 1031915, "author_name": "stephenliang", "author_url": "", "post_date": "09/29/2020 19:18:56", "content": "https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet
", "votes": null, "replies": [] }, { "id": 1032054, "author_name": "hitman47", "author_url": "", "post_date": "09/29/2020 23:00:16", "content": "Hi. I am using pretrained models from repo https://github.com/lfz/DSB2017 which has open source license that was mentioned in the dsb2017 competition rules; cite of paper arXiv:1711.08324 will be done.
", "votes": null, "replies": [] }, { "id": 1035642, "author_name": "dysonlin", "author_url": "", "post_date": "10/02/2020 23:01:49", "content": "https://github.com/facebook/prophet
", "votes": null, "replies": [] }, { "id": 921754, "author_name": "hojijoji", "author_url": "", "post_date": "07/09/2020 14:16:16", "content": "Can models for pre-processing (e.g. segmentation) be used if they were trained on private data? Is it required that such a model is declared and made publicly available prior to the entry deadline?
", "votes": null, "replies": [ { "id": 921905, "author_name": "juliaelliott", "author_url": "", "post_date": "07/09/2020 16:36:03", "content": "@hojijoji If the pre-trained model itself is freely and publicly licensed for \"use that includes research and academic purposes,\" and the resulting solution/code could be open source licensed (including for commercial use) if it in a winning position, then yes, it can be used. However, if licensing prevents these requirements from being met, then it cannot be used. And yes, the pre-trained model should be declared prior to the entry deadline.
", "votes": null, "replies": [ { "id": 967776, "author_name": "kurianbenoy", "author_url": "", "post_date": "08/12/2020 13:55:32", "content": "Is open source licensing okay for you?
", "votes": null, "replies": [] }, { "id": 968035, "author_name": "juliaelliott", "author_url": "", "post_date": "08/12/2020 16:42:32", "content": "@kurianbenoy - Yes, if external data is open source licensed and freely & publicly available (no cost to any other competitors), then it is permitted. You would need to declare the external data you are using on this forum thread before the entry deadline.
", "votes": null, "replies": [] } ] } ] }, { "id": 925231, "author_name": "ronaldokun", "author_url": "", "post_date": "07/11/2020 23:56:56", "content": "The Dataset http://medgift.hevs.ch/wordpress/databases/ild-database/, it has the the following description:\n> The database is available for research on request and after signature of a license agreement. \nyou have to sign and send the licensed agreement by email then you can access the data.
\n\nAlthough it's not readily available, it is so by request for research and academic purposes.
\n\n@juliaelliott Can you confirm it if it abides by the competition rules?
\n\nThank you in advance
", "votes": null, "replies": [ { "id": 930721, "author_name": "ronaldokun", "author_url": "", "post_date": "07/15/2020 17:07:26", "content": "Well, by consulting the licensed agreement we would have to sign in the website above, it says:\n> The End-User named must be a legal institution, or a department or section of a named legal institution, not an individual nor a project -the person signing this Agreement has to be duly authorized by the institution for such signatures (e.g., Department or Administrative Head or similar) and shall be liable for such authorization.
\n\nSince we Kagglers would be classified as persons and not a Research Institution, I didn't bother to request access to the Database mentioned and abandoned the effort. So the Dataset will not be used for me and the message can be ignored.
\n\nIf anyone else succeeds, please let me know.
", "votes": null, "replies": [] }, { "id": 930743, "author_name": "juliaelliott", "author_url": "", "post_date": "07/15/2020 17:21:05", "content": "@ronaldokun Based on your observation regarding a gating mechanism on the dataset which would make its access and availability not fully public to all other competitors, this would be prohibited by the rules. If, however, there are research-use-licensed datasets which are not gated (and therefore freely/publicly available to all competitors), the host has specified that such datasets are permitted, so long as the resulting solution/code being created, if in winning standing, will be licensed as open source, including commercial use.
", "votes": null, "replies": [] }, { "id": 930745, "author_name": "ronaldokun", "author_url": "", "post_date": "07/15/2020 17:22:26", "content": "Thank you Julia for the clarification.
", "votes": null, "replies": [] } ] }, { "id": 944273, "author_name": "jjinho", "author_url": "", "post_date": "07/25/2020 01:47:19", "content": "Computed Tomography Emphysema Database
\n\nLink: https://lauge-soerensen.github.io/emphysema-database/
\n\nAs per authors is free of charge for research and educational purposes. Uses TIFF.
", "votes": null, "replies": [ { "id": 944289, "author_name": "rpsantosakaggle", "author_url": "", "post_date": "07/25/2020 02:02:09", "content": "Emphysema is a different disease
", "votes": null, "replies": [] }, { "id": 944359, "author_name": "jjinho", "author_url": "", "post_date": "07/25/2020 03:44:40", "content": "That is true. But there are overlaps with the two diseases, including an entity called \"Combined Pulmonary Fibrosis and Emphysema Syndrome.\"
\n\n\n\n\nRadiographic characteristics include the presence of upper-lobe emphysema and lower-lobe pulmonary fibrosis.
\n
This paper also notes that the proportion of patients with combined PF and emphysema can range from 18.8% to 50.9%. It seems probable that a proportion of the patients will possibly have this component as well. Note that these patients are said to experience slower decline in FVC (which elucidating is a clear goal of this competition). Either way, it seems having the ability to identify radiologic findings consistent with emphysema and then have that as a feature may be helpful given what the literature states. YMMV
", "votes": null, "replies": [] } ] }, { "id": 944693, "author_name": "aidcostello", "author_url": "", "post_date": "07/25/2020 09:02:14", "content": "I am using lungmask, licensed under the GNU Public License
\n\nJohannes Hofmanninger, Forian Prayer, Jeanny Pan, Sebastian Röhrich, Helmut Prosch and Georg Langs. \"Automatic lung segmentation in routine imaging is a data diversity problem, not a methodology problem\". 1 2020, https://arxiv.org/abs/2001.11767
", "votes": null, "replies": [ { "id": 955980, "author_name": "bluesummers", "author_url": "", "post_date": "08/03/2020 06:06:48", "content": "Using the same, to be specific to an implementation (as it is considered data) - https://github.com/JoHof/lungmask
", "votes": null, "replies": [] }, { "id": 957705, "author_name": "melvin97n", "author_url": "", "post_date": "08/04/2020 13:58:13", "content": "Have you or anyone run this on Kaggle without internet? do respond
", "votes": null, "replies": [] }, { "id": 991757, "author_name": "gautham11", "author_url": "", "post_date": "08/30/2020 16:35:38", "content": "I did run this without the internet, by creating a dataset. I compared it with other segmentation methods in this kernel
\nhttps://www.kaggle.com/gautham11/lung-segmentation-methods-comparison
\nBut I can't use this in any submission, because it takes ~2 seconds to segment an image. Let me know if anyone managed to get around this.
\nWatershed segmentation seems like the way to go.
About (2.) - I want to create this offline, as it takes long time, upload it somewhere to make it public and share it with others so it will be public and usable
", "votes": null, "replies": [ { "id": 955940, "author_name": "matthewmasters", "author_url": "", "post_date": "08/03/2020 05:20:56", "content": "I was wondering: are we allowed to train our models on our own machines, and then simply let the notebook do the predictions?
\n\nLet's say I wanted to use an auto-encoder to handle the images, and then use another prediction model to combine the data from the images, with the FVC data, could I train the autoencoder on my own machine, then add the trained model as a resource and load it in my notebook to do the rest of the pipeline?\nOr does any training with the actual data have to happen in the notebook within those time constraints?
\n\nThank you
", "votes": null, "replies": [ { "id": 966862, "author_name": "juliaelliott", "author_url": "", "post_date": "08/11/2020 17:41:47", "content": "@stadlerm Yes, you can train locally and upload the trained model into Kaggle's notebook as an external data source, effectively requiring that you minimally run inference in your notebook submission on Kaggle. The code must be possible to re-run on a hidden test set swapped out for the current test set.
", "votes": null, "replies": [] }, { "id": 1011058, "author_name": "ahmedewais", "author_url": "", "post_date": "09/15/2020 08:17:14", "content": "if I train locally and then upload the weights, do I have to share my weights here as external data source or make it open source?
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "919240": "Per the [Competition Rules](https://www.kaggle.com/c/osic-pulmonary-fibrosis-progression/rules), **publicly and freely available external data is permitted, if it is available for use that includes research or academic purposes**. If you are a winner, your solution must also be released under the stated open source license, per the rules' winner license requirements. \n\nAll external data must be posted to this forum thread no later than the Entry Deadline (one week before competition close).\n\nOnce someone posts an external dataset to this thread, you do not need to re-post it if you are using the same one.\n\nYou only need to declare the original dataset used; you do not need to declare re-labeled or augmented or otherwise processed versions of datasets. Pre-trained models can be declared, as well; however, models resulting from your own original work, that you have trained yourself offline do not need to be shared/declared.", "921754": "Can models for pre-processing (e.g. segmentation) be used if they were trained on private data? Is it required that such a model is declared and made publicly available prior to the entry deadline?", "921905": "hojijoji If the pre-trained model itself is freely and publicly licensed for \"use that includes research and academic purposes,\" and the resulting solution/code could be open source licensed (including for commercial use) if it in a winning position, then yes, it can be used. However, if licensing prevents these requirements from being met, then it cannot be used. And yes, the pre-trained model should be declared prior to the entry deadline.", "925231": "The Dataset http://medgift.hevs.ch/wordpress/databases/ild-database/, it has the the following description:\n> The database is available for research on request and after signature of a license agreement. \nyou have to sign and send the licensed agreement by email then you can access the data.\n\nAlthough it's not readily available, it is so by request for research and academic purposes.\n\n@juliaelliott Can you confirm it if it abides by the competition rules?\n\nThank you in advance", "930721": "Well, by consulting the licensed agreement we would have to sign in the website above, it says:\n> The End-User named must be a legal institution, or a department or section of a named legal institution, not an individual nor a project -the person signing this Agreement has to be duly authorized by the institution for such signatures (e.g., Department or Administrative Head or similar) and shall be liable for such authorization.\n\nSince we Kagglers would be classified as persons and not a Research Institution, I didn't bother to request access to the Database mentioned and abandoned the effort. So the Dataset will **not** be used for me and the message can be ignored.\n\nIf anyone else succeeds, please let me know.", "930743": "ronaldokun Based on your observation regarding a gating mechanism on the dataset which would make its access and availability not fully public to all other competitors, this would be prohibited by the rules. If, however, there are research-use-licensed datasets which are not gated (and therefore freely/publicly available to all competitors), the host has specified that such datasets are permitted, so long as the resulting solution/code being created, if in winning standing, will be licensed as open source, including commercial use.", "930745": "Thank you Julia for the clarification.", "944273": "**Computed Tomography Emphysema Database**\n\nLink: https://lauge-soerensen.github.io/emphysema-database/\n\nAs per authors is free of charge for research and educational purposes. Uses TIFF.", "944289": "Emphysema is a different disease", "944359": "That is true. But there are overlaps with the two diseases, including an entity called \"[Combined Pulmonary Fibrosis and Emphysema Syndrome](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3251269/).\" \n\n> Radiographic characteristics include the presence of upper-lobe emphysema and lower-lobe pulmonary fibrosis.\n\nThis paper also notes that the proportion of patients with combined PF and emphysema can range from 18.8% to 50.9%. It seems probable that a proportion of the patients will possibly have this component as well. Note that these patients are said to experience slower decline in FVC (which elucidating is a clear goal of this competition). Either way, it seems having the ability to identify radiologic findings consistent with emphysema and then have that as a feature may be helpful given what the literature states. YMMV", "944693": "I am using lungmask, licensed under the GNU Public License\n\nJohannes Hofmanninger, Forian Prayer, Jeanny Pan, Sebastian Röhrich, Helmut Prosch and Georg Langs. \"Automatic lung segmentation in routine imaging is a data diversity problem, not a methodology problem\". 1 2020, https://arxiv.org/abs/2001.11767", "955933": "1. If kernels have no internet access, how am I to use external data whether of third party or my own?\n2. If I create my own dataset (segmented version of the images) and I want to share it publicly (so we all can use it) - what is the procedure to do so?\n\nAbout (2.) - I want to create this offline, as it takes long time, upload it somewhere to make it public and share it with others so it will be public and usable", "955940": "1. They have internet access disabled but you can still add Kaggle datasets as input. So any external data, weights, etc. can be uploaded as a public or private Kaggle dataset and imported.\n2. You can just use a public Kaggle dataset for this. If you do make sure to add the appropriate license, ensure it's compatible with the competition's licence, and attribute the original creator.", "955980": "Using the same, to be specific to an implementation (as it is considered data) - https://github.com/JoHof/lungmask", "957705": "Have you or anyone run this on Kaggle without internet? do respond", "966270": "I was wondering: are we allowed to train our models on our own machines, and then simply let the notebook do the predictions?\n\nLet's say I wanted to use an auto-encoder to handle the images, and then use another prediction model to combine the data from the images, with the FVC data, could I train the autoencoder on my own machine, then add the trained model as a resource and load it in my notebook to do the rest of the pipeline?\nOr does any training with the actual data have to happen in the notebook within those time constraints?\n\nThank you", "966862": "stadlerm Yes, you can train locally and upload the trained model into Kaggle's notebook as an external data source, effectively requiring that you minimally run inference in your notebook submission on Kaggle. The code must be possible to re-run on a hidden test set swapped out for the current test set.", "967776": "Is open source licensing okay for you?", "968035": "kurianbenoy - Yes, if external data is open source licensed and freely & publicly available (no cost to any other competitors), then it is permitted. You would need to declare the external data you are using on this forum thread before the entry deadline.", "991757": "I did run this without the internet, by creating a dataset. I compared it with other segmentation methods in this kernel\n[https://www.kaggle.com/gautham11/lung-segmentation-methods-comparison](https://www.kaggle.com/gautham11/lung-segmentation-methods-comparison)\nBut I can't use this in any submission, because it takes ~2 seconds to segment an image. Let me know if anyone managed to get around this.\nWatershed segmentation seems like the way to go.", "995399": "Are these licenses allowed in this competition? **(CC BY NC 3.0, CC BY NC SA 3.0, CC BY 4.0)**", "995586": "orkatz2 The competition permits publicly and freely available external data is permitted if it is available for use that includes research or academic purposes. These all appear to be licenses that are available for research/academic purposes, however you should also ensure they are free to all and publicly available.", "997419": "A model related to segmentation\n\nHere is an interesting link to the segmentation of biological images including Lung using the BCDU network.\nhttps://github.com/rezazad68/BCDU-Net\nIt seems it is freely available according to the author, He only mentioned CITE them. I haven't found any Licence file in the GitHub. May I use them?", "998236": "nishyanand If you cannot properly identify provenance of and licensing of the dataset, I would recommend not using it.", "998651": "I have sent an email to the author(rezazad68@gmail.com is the email address of the first author mentioned in the GitHub link) and got a response as\n\"*Code and weight are open access just cite the paper in case of using them*.\".\n I have attached the email conversation as well. I hope this is enough for use it. @juliaelliott please let me know.", "1003083": "nishyanand This sounds acceptable.", "1009306": "I have found that there are several R packages that are in CRAN but for some reason are missing from the notebook R script environment. But I am able to upload them via \"Add data\" as a zip file and then execute the commands in my script to unpack and install them. Do I need to declare these \"datasets\" (zip files of R packages) as external data in this thread?", "1010483": "lungmask https://github.com/JoHof/lungmask", "1011058": "if I train locally and then upload the weights, do I have to share my weights here as external data source or make it open source?", "1015668": "Can i use the DataSet learning from this contest for a Research paper by citing the Data Source?", "1015998": "siddarthpaim Yes, you may use the dataset in compliance with the [data use terms (Section 7) in the competition rules](https://www.kaggle.com/c/osic-pulmonary-fibrosis-progression/rules).", "1016003": "dslate Packages are not necessary to declare. Pre-trained models are.", "1027533": "(https://github.com/JoHof/lungmask)\nAutomated lung segmentation in CT", "1028362": "Pretrained models from: \nhttps://pytorch.org/docs/stable/torchvision/models.html\nhttps://github.com/rwightman/pytorch-image-models/", "1030026": "I am using below pretrained model\nhttps://www.kaggle.com/khoongweihao/osic-model-weights", "1030191": "Hi @juliaelliott , is GPLv3 allowed? I read on thread [here](https://www.kaggle.com/c/osic-pulmonary-fibrosis-progression/discussion/180026), which looks like it is, commercial and privates are both permitted, but I didn't see this license being mentioned anywhere else in the forum, so just want to confirm it. Thanks.", "1030469": "samshipengs \nIs it publicly available at no cost to other participants?\nIs it licensed for use that includes academic/research purposes?\nIf you use the dataset, is it possible for you to license your solution/code in accordance with the competition's \"Winners License\" rules?\n\nIf the answer to those 3 questions is yes, then yes, it is permitted. If any answer is no, then no, it is not permitted.", "1030598": "samshipengs actually I looked at the link in that thread in more detail and found something. You can see my response [here](https://www.kaggle.com/c/osic-pulmonary-fibrosis-progression/discussion/180026#1030594).\n\n@juliaelliott could you please have a quick look also? I understand that your three question system is meant to make things simple, but that third question:\n\n> If you use the dataset, is it possible for you to license your solution/code in accordance with the competition's \"Winners License\" rules?\n\nis actually really involved in terms of understanding copyright laws. Wondering if there is anyone on your team who can definitively tell us whether we may use and/or adapt GPLv3 licensed code.", "1030614": "https://www.kaggle.com/andrewmvd/covid19-ct-scans", "1030660": "alexandersoare Thanks for your question in that thread, actually that part was exactly what I didnt understand when I was reading the link shared by @stanleyjzheng , so I though its better to check with officials. just for reference, this is the part i didnt understand:\n> Note that compatibility with the GPLv3 is one-way only, which means you may license your contributions to adaptations of BY-SA 4.0 materials under GPLv3, but you may not license your contributions to adaptations of GPLv3 projects under BY-SA 4.0. Other special considerations apply. See the full analysis and comparison for more information. \n\nMy attempt at this is, it means you can use GPLv3 code for BY-SA4 projects but not the other way around (i.e. you can't use BY-SA4 code for GPLv3 projects), is that correct? \n\n@juliaelliott thanks for your guide on checking allowed datasets, I think the answer to the first two questions are affirmative, its no cost for everyone to use and i suppose there is no issue to use for academic/research since its even okay for commercial use, but im really not sure for the third question. Would really appreciate if you could provide more clarity.", "1030692": "samshipengs If using that dataset prohibits you to license your final solution as required by the competition rules, then the 3rd question I posed would be no, and that would make it prohibited for use.", "1030695": "Thanks for the reply, although the issue is I'm not sure if that data set would prohibit me to license my solution as required by the rules (I referenced above) and hoping to get clarification. Is it just me not understanding it? if so could someone help to explain?\n\n@hojijoji Would you share some insights if you don't mind? (I believe you are the author of the repo, thanks for the great work!)", "1030731": "juliaelliott may we use the scans currently available in the [RSNA STR Pulmonary Embolism Detection\n](https://www.kaggle.com/c/rsna-str-pulmonary-embolism-detection) competition? \n\nParticularly for unsupervised learning", "1030743": "https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf2_detection_zoo.md\nhttps://github.com/UCSD-AI4H/COVID-CT \nhttps://github.com/JoHof/lungmask\nhttp://www.via.cornell.edu/crpf.html\nhttps://www.kaggle.com/andrewmvd/covid19-ct-scans\nhttps://wiki.cancerimagingarchive.net/display/Public/NSCLC+Radiogenomics\nhttps://wiki.cancerimagingarchive.net/display/Public/RIDER+Lung+CT\nhttps://wiki.cancerimagingarchive.net/display/Public/LIDC-IDRI\nhttps://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=70224216\nhttps://wiki.cancerimagingarchive.net/display/Public/Lung+Fused-CT-Pathology\nhttps://wiki.cancerimagingarchive.net/display/Public/NSCLC-Radiomics\nhttps://wiki.cancerimagingarchive.net/display/Public/NSCLC-Radiomics-Interobserver1\nhttps://wiki.cancerimagingarchive.net/display/Public/TCGA-LUSC\nhttps://wiki.cancerimagingarchive.net/display/Public/LungCT-Diagnosis\nhttps://wiki.cancerimagingarchive.net/display/Public/NSCLC-Radiomics-Genomics\nhttps://wiki.cancerimagingarchive.net/display/Public/SPIE-AAPM+Lung+CT+Challenge\nhttps://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=52758026\nhttps://wiki.cancerimagingarchive.net/display/Public/LIDC-IDRI", "1030749": "https://github.com/shijianjian/EfficientNet-PyTorch-3D", "1030750": "tannercarbonati That competition's dataset is licensed for non-commercial purposes, so it should be eligible to use as long as you are not redistributing it.", "1031117": "juliaelliott just to clarify this is not about datasets (from my end at least). It's about public code with a GPLv3 license. Surely someone will need to check when it comes time to evaluate the final solutions, so any chance we can get a clear answer now?", "1031125": "My understanding is that you can not relicense GPLv3 (copyleft) material under CC-BY 4.0 (permissive). Also, you should not confuse CC-BY (permissive) with CC-BY-SA (copyleft). However, eventhough I would have prefered to stick to a copyleft license, I have changed the [repository ](https://github.com/JoHof/lungmask) to Apache 2.0 which should be ok with CC-BY 4.0 so that you can use it for this competition. Note, that this only affects to current version!", "1031190": "hojijoji many thanks!", "1031301": "thanks for the explanation and modifying license for current version!", "1031422": "EfficientNet B0 to B7\nhttps://github.com/qubvel/efficientnet", "1031582": "https://github.com/qubvel/efficientnet\nhttps://github.com/JoHof/lungmask/\nhttps://github.com/dlpbc/keras-kinetics-i3d", "1031822": "https://www.kaggle.com/bjaeger/pretrained-weights", "1031915": "https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet", "1032054": "Hi. I am using pretrained models from repo https://github.com/lfz/DSB2017 which has open source license that was mentioned in the dsb2017 competition rules; cite of paper arXiv:1711.08324 will be done.", "1035642": "https://github.com/facebook/prophet" }, "source": "meta" }