\n
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F10119745%2F00666b491fd576af05cab7dcf5fc58d6%2F2.png?generation=1679537131570729&alt=media\")
\nBy the way, I have a question. I ran and submitted the Vesuvius Challenge: Example submission file in the discussion area, but it keeps showing \"notebook timeout.\" Has anyone else encountered this issue?
\n
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F10119745%2F7e287701f51b4ea5715568d9f56b7dcd%2F1.png?generation=1679537144598141&alt=media\")
Hey, welcome to the Ink Detection Progress Prize of the Vesuvius Challenge!
\nYou might find the following resources particularly useful:
\nFeel free to ask any and all questions. We're here to help out. Good luck and have fun!
\nโ Daniel & JP
Hey, welcome to the Ink Detection Progress Prize of the Vesuvius Challenge!
\nYou might find the following resources particularly useful:
\nFeel free to ask any and all questions. We're here to help out. Good luck and have fun!
\nโ Daniel & JP
Hello, @jpposma, in previous competition at kaggle HuBMAP + HPA - Hacking the Human Body hosts answered to the participant that they are a non-profit organization, so it was one of reasons why SegFormer was admissible with constrained usage licence SegFormer/LICENSE in the competition.
\nSorry, if this is obvious, I don't have enough experience with model licenses for their usage at kaggle yet. Can I use this model in this competition too? Maybe this question will be useful for other people
", "rawMarkdown": "Hello, @jpposma, in previous competition at kaggle [HuBMAP + HPA - Hacking the Human Body](https://www.kaggle.com/competitions/hubmap-organ-segmentation/overview) hosts answered to the participant that they are a [non-profit organization] (https://www.kaggle.com/competitions/hubmap-organ-segmentation/discussion/356201), so it was one of reasons why SegFormer was admissible with constrained usage licence [SegFormer/LICENSE](https://github.com/NVlabs/SegFormer/blob/master/LICENSE) in the competition.\n\nSorry, if this is obvious, I don't have enough experience with model licenses for their usage at kaggle yet. Can I use this model in this competition too? Maybe this question will be useful for other people", "votes": 2, "replies": [ { "id": 2272765, "postDate": "2023-05-24T18:03:34.570Z", "content": "For anyone coming across this, we clarified here (TL;DR: yes you can use those) https://www.kaggle.com/competitions/vesuvius-challenge-ink-detection/discussion/410906
", "rawMarkdown": "For anyone coming across this, we clarified here (TL;DR: yes you can use those) https://www.kaggle.com/competitions/vesuvius-challenge-ink-detection/discussion/410906" } ] }, { "id": 2185362, "postDate": "2023-03-17T02:00:50.567Z", "content": "This is very cool.
", "rawMarkdown": "This is very cool.", "votes": 4, "replies": [ { "id": 2185380, "postDate": "2023-03-17T02:36:08.217Z", "content": "@hallayang it's been a while since seeing you around the forums/leaderboards. Welcome back! :)
", "rawMarkdown": "@hallayang it's been a while since seeing you around the forums/leaderboards. Welcome back! :)", "votes": 2 } ] }, { "id": 2191587, "postDate": "2023-03-22T04:35:12.423Z", "content": "G'day Kaggle gods! Thanks for this super interesting competition!
\nI come asking stupid questions. ๐
\nIt's been 11 years since I've entered a Kaggle comp, and that one got me a job, a job I still have today. (Excluding, I guess, the two later Facebook Recruiting Competitions that I was on the admin team for, which I hope doesn't disqualify me from entering one as a competitor.) (And I hope that having lunch with you guys down at Kaggle HQ near the ball park right after I started at Facebook doesn't disqualify me either.) ๐
\nMy quandary is this: for that 2012 comp I was making up my own ML system. It was in C#, and I had to submit a copy of my code and my documentation and a license for Facebook to use it forever everywhere all at once with my final submission. But before that I could just submit submissions.
\nLong story short, I only just finished building that ML system. (Yeah, not very \"move fast\"; don't get me started.) ๐ Anyway, I was looking to see if there were any new comps to test it out on, and this one looks perfect. And you just kicked it off a week ago.
\nThe problem is that you've made a lot of improvements to Kaggle in the past 11 years. (Duh. What did I expect?) And it looks like it's only possible to make submissions if you do it in a Notebook in Python or R, maybe pulling in some standard AMLT.
\nIs there any way for me to make submissions based on Bear?
\nIf not, that's fine. I only wanted to test out Bear, and even on the off chance that I won anything (which I didn't do last time, although I did get the job, I suppose) I would donate it to some bear-related charity anyway.
\nIn any case, I have the data now (ha, rubbing hands in glee whenever I get a super interesting dataset) ๐ and I'd be happy to report anecdotally whether Bear digests it well or not.
\nThanks in advance!
", "rawMarkdown": "G'day Kaggle gods! Thanks for this super interesting competition!\n\nI come asking stupid questions. ๐\n\nIt's been 11 years since I've entered a Kaggle comp, and [that one got me a job](https://johncostella.com/bear/history/), a job I still have today. (Excluding, I guess, the two later Facebook Recruiting Competitions that I was on the admin team for, which I hope doesn't disqualify me from entering one as a competitor.) (And I hope that having lunch with you guys down at Kaggle HQ near the ball park right after I started at Facebook doesn't disqualify me either.) ๐\n\nMy quandary is this: for that 2012 comp I was making up my own ML system. It was in C#, and I had to submit a copy of my code and my documentation and a license for Facebook to use it forever everywhere all at once with my final submission. But before that I could just submit submissions.\n\nLong story short, [I only just finished building that ML system](https://johncostella.com/bear/). (Yeah, not very \"move fast\"; don't get me started.) ๐ Anyway, I was looking to see if there were any new comps to test it out on, and this one looks perfect. And you just kicked it off a week ago.\n\nThe problem is that you've made a lot of improvements to Kaggle in the past 11 years. (Duh. What did I expect?) And it looks like it's only possible to make submissions if you do it in a Notebook in Python or R, maybe pulling in some standard AMLT.\n\nIs there any way for me to make submissions based on Bear?\n\nIf not, that's fine. I only wanted to test out Bear, and even on the off chance that I won anything (which I didn't do last time, although I did get the job, I suppose) I would donate it to some bear-related charity anyway.\n\nIn any case, I have the data now (ha, rubbing hands in glee whenever I get a super interesting dataset) ๐ and I'd be happy to report anecdotally whether Bear digests it well or not.\n\nThanks in advance!", "votes": 1, "replies": [ { "id": 2192334, "postDate": "2023-03-22T15:39:10.417Z", "content": "Welcome back to Kaggle! Sure, you can use whatever you want. Good luck!
", "rawMarkdown": "Welcome back to Kaggle! Sure, you can use whatever you want. Good luck!", "votes": 1, "replies": [ { "id": 2192347, "postDate": "2023-03-22T15:49:40.063Z", "content": "But the submissions require use of a Notebook for the code, not just a submission CSV file, right? Or am I misunderstanding the restrictions here?
", "rawMarkdown": "But the submissions require use of a Notebook for the code, not just a submission CSV file, right? Or am I misunderstanding the restrictions here?" }, { "id": 2192351, "postDate": "2023-03-22T15:51:25.650Z", "content": "Correct. But you can run within that notebook whatever you want.
", "rawMarkdown": "Correct. But you can run within that notebook whatever you want.", "votes": 1 }, { "id": 2192383, "postDate": "2023-03-22T16:13:07.320Z", "content": "๐ Ah. Got it. I didn't think that that was legit, but I see that it fits within the comp rules.
\nThanks!
", "rawMarkdown": "๐ Ah. Got it. I didn't think that that was legit, but I see that it fits within the comp rules.\n\nThanks!" }, { "id": 2193013, "postDate": "2023-03-23T03:38:22.730Z", "content": "OK, so now that I actually read the Data text properly (see my \"Dumb question about the fragments\" question), I'm confused again.
\nIf the test fragment masks and CT scans are hidden (substituted by mocks in the supplied dataโthat's what I missed when I read it the first few times), then the Notebook has to shell out to Bear itself. But the Notebook is hosted on your end, right? How can I compile a program on that host so that the Notebook can shell out to it and run it?
", "rawMarkdown": "OK, so now that I actually read the Data text properly (see my \"Dumb question about the fragments\" question), I'm confused again.\n\nIf the test fragment masks and CT scans are hidden (substituted by mocks in the supplied dataโthat's what I missed when I read it the first few times), then the Notebook has to shell out to Bear itself. But the Notebook is hosted on your end, right? How can I compile a program on that host so that the Notebook can shell out to it and run it?" } ] } ] }, { "id": 2183856, "postDate": "2023-03-16T01:03:54Z", "content": "Hi, thanks for organizing an incredible competition.
\n\" There is a $150,000 grand prize available to the first team that can read these scrolls from a 3d x-ray scan. \"
\nhttps://scrollprize.org/participate
I suppose this is another separate competition. Can this solution uses winning solution from kaggle ink detection?
", "rawMarkdown": "Hi, thanks for organizing an incredible competition.\n\n\" There is a $150,000 grand prize available to the first team that can read these scrolls from a 3d x-ray scan. \" \nhttps://scrollprize.org/participate\n\nI suppose this is another separate competition. Can this solution uses winning solution from kaggle ink detection?", "votes": 1, "replies": [ { "id": 2183857, "postDate": "2023-03-16T01:05:54.650Z", "content": "Yes, for that separate competition you can use whatever method that you want. When Ink Detection finishes the winning solutions will be made open source, so you can use those then.
", "rawMarkdown": "Yes, for that separate competition you can use whatever method that you want. When Ink Detection finishes the winning solutions will be made open source, so you can use those then." } ] }, { "id": 2264336, "postDate": "2023-05-18T10:49:18.933Z", "content": "Hi team,
\nI have just started learning ML/DL and all. And this is my first attempt on Kaggle competition, to get a brief understanding of how it goes and how the competitions are like.
\nI am trying to search how should I approach, what model should I use, etc..
\nBut the thing is I have no clue with what keyword I should search with.
\nI tried to search ink detection, image detection, 3D image classification, these didn't work out well.
\nCould anyone help me with understanding the concept and find if there is any stackoverflow queries, youtube videos, or any material where I can learn more about it?
\nThanks
I think this competition is highly meaningful. Although my knowledge and experience may not be sufficient, I would like to see the solutions from everyone else.
", "rawMarkdown": "I think this competition is highly meaningful. Although my knowledge and experience may not be sufficient, I would like to see the solutions from everyone else." }, { "id": 2256320, "postDate": "2023-05-12T11:42:33.180Z", "content": "Hi @danielhavir @jpposma it is very frustrating that we still do not have answers about this topic https://www.kaggle.com/competitions/vesuvius-challenge-ink-detection/discussion/407355- can you give us an answer or at least give some estimate when you will decide what to do with segformer models (for me it look that competition is non profit and we can use any models)
", "rawMarkdown": "Hi @danielhavir @jpposma it is very frustrating that we still do not have answers about this topic https://www.kaggle.com/competitions/vesuvius-challenge-ink-detection/discussion/407355- can you give us an answer or at least give some estimate when you will decide what to do with segformer models (for me it look that competition is non profit and we can use any models)" }, { "id": 2254110, "postDate": "2023-05-10T16:55:31.293Z", "content": "Anyone here.. Who hires the data scientist in his/her team?
\nI have good knowledge about data science but I have no experience about this industry further I am really excited to work as a data scientist.
\nI am a second year student of computer science and I have a diploma of data science with NLP domain.
\nI do not need money I need some experience.
Great Competition!
", "rawMarkdown": "Great Competition!" }, { "id": 2243952, "postDate": "2023-05-03T10:13:11.487Z", "content": "Thank you for this exciting challenge and for the example submission notebook.
\nHowever I'm unable to save&commit the notebook while using the generated model weights.
\nApologies for cross-posting - I asked this question here (https://www.kaggle.com/competitions/vesuvius-challenge-ink-detection/discussion/406421 ) but didn't get a response. Thought I'd have better chances of getting response here.
\nThanks in advance
As I could understand, there are some channels / discussion on Discord. How do I connect these channels?
", "rawMarkdown": "As I could understand, there are some channels / discussion on Discord. How do I connect these channels?" }, { "id": 2233965, "postDate": "2023-04-24T18:16:43.217Z", "content": "This is so cool!
", "rawMarkdown": "This is so cool!" }, { "id": 2229518, "postDate": "2023-04-21T12:49:55.633Z", "content": "It's really interesting, although i haven't submitted any of my work till now , but really looking forward to make this one my very first one to the path of gloryโฆ
", "rawMarkdown": "It's really interesting, although i haven't submitted any of my work till now , but really looking forward to make this one my very first one to the path of glory... " }, { "id": 2199647, "postDate": "2023-03-27T22:37:23.243Z", "content": "Interesting competition!
", "rawMarkdown": "Interesting competition!" }, { "id": 2199340, "postDate": "2023-03-27T16:37:25.697Z", "content": "Advice for beginner in Machine learning
", "rawMarkdown": "Advice for beginner in Machine learning\n" }, { "id": 2192641, "postDate": "2023-03-22T19:45:21.603Z", "content": "Just some clarification, hosts:
\nI am unable to submit / output ?
", "rawMarkdown": "I am unable to submit / output ? " }, { "id": 2190809, "postDate": "2023-03-21T14:05:35.373Z", "content": "Hey Everyone new to the site / once I submit and run my model and it has completed where can I find status etcโฆ. not seeing myself on leaderboard ?
", "rawMarkdown": "Hey Everyone new to the site / once I submit and run my model and it has completed where can I find status etc.... not seeing myself on leaderboard ? ", "replies": [ { "id": 2192936, "postDate": "2023-03-23T02:03:12.923Z", "content": "You can find the submission button on the right-hand side of the Ink Detection competition page. After clicking it, you can see the status of your submission.
\n
\nBy the way, I have a question. I ran and submitted the Vesuvius Challenge: Example submission file in the discussion area, but it keeps showing \"notebook timeout.\" Has anyone else encountered this issue?
\n
I had the same issue. Not sure I have seen anyone share a successful submission of this example. For sure you will need to not train and predict in the same session. So use the provided model or train your own, load it and run just the prediction.
\nI am trying to re-write and re-train using the dual GPU that kaggle provides. That might be fast enough to beat the clock. But getting torch to use two GPU's is often harder than it seems it should be :)
\nSince I have dual GPU on my local machines I am going to follow that path - but you might also play with larger batch sizing to see if you can beat the clock without hitting the memory limit on kaggle.
", "rawMarkdown": "I had the same issue. Not sure I have seen anyone share a successful submission of this example. For sure you will need to not train and predict in the same session. So use the provided model or train your own, load it and run just the prediction.\n\nI am trying to re-write and re-train using the dual GPU that kaggle provides. That might be fast enough to beat the clock. But getting torch to use two GPU's is often harder than it seems it should be :)\n\nSince I have dual GPU on my local machines I am going to follow that path - but you might also play with larger batch sizing to see if you can beat the clock without hitting the memory limit on kaggle." }, { "id": 2193204, "postDate": "2023-03-23T06:23:10.780Z", "content": "Thanks for your suggestion, I will try it later :)
", "rawMarkdown": "Thanks for your suggestion, I will try it later :)" } ] }, { "id": 2195524, "postDate": "2023-03-24T17:39:21.243Z", "content": "There are some submission rules that you may not obey.
", "rawMarkdown": "There are some submission rules that you may not obey." } ] }, { "id": 2189924, "postDate": "2023-03-20T22:27:56.083Z", "content": "very good !
", "rawMarkdown": "very good !" }, { "id": 2189856, "postDate": "2023-03-20T20:52:43.747Z", "content": "very nice
", "rawMarkdown": "very nice\n" }, { "id": 2189449, "postDate": "2023-03-20T14:05:22.267Z", "content": "very cool
", "rawMarkdown": "very cool\n" }, { "id": 2185984, "postDate": "2023-03-17T13:21:41.430Z", "content": "\n\nYou can expect two fragments in the hidden test set, which together are roughly the same size as a single training fragment.
\n
So is the private LB scored on only 1 fragment? My concern is that it may lead to solutions which overfit to this one fragment score higher on the final LB than more generalizable solutions.
", "rawMarkdown": "> You can expect two fragments in the hidden test set, which together are roughly the same size as a single training fragment.\n\nSo is the private LB scored on only **1** fragment? My concern is that it may lead to solutions which overfit to this one fragment score higher on the final LB than more generalizable solutions.", "replies": [ { "id": 2186151, "postDate": "2023-03-17T15:48:33.993Z", "content": "We only had 4 scanned fragments in total, so we didn't have much of a choice, unfortunately. We would have loved more data too, but this is what we have!
", "rawMarkdown": "We only had 4 scanned fragments in total, so we didn't have much of a choice, unfortunately. We would have loved more data too, but this is what we have!", "votes": 2 } ] }, { "id": 2185907, "postDate": "2023-03-17T12:01:52.220Z", "content": "What an interesting competition! Like it is something mystical to learn about this manuscripts from the past. I will invest my time for sure. Thanks host!
", "rawMarkdown": "What an interesting competition! Like it is something mystical to learn about this manuscripts from the past. I will invest my time for sure. Thanks host!", "replies": [ { "id": 2186149, "postDate": "2023-03-17T15:47:55.413Z", "content": "Thank you :)
", "rawMarkdown": "Thank you :)" } ] }, { "id": 2184423, "postDate": "2023-03-16T11:25:58.893Z", "content": "Please take me along while you loop for Vesuvius coding project. I like to see to the end of all its going to take.
", "rawMarkdown": "Please take me along while you loop for Vesuvius coding project. I like to see to the end of all its going to take.", "replies": [ { "id": 2184769, "postDate": "2023-03-16T15:54:10.423Z", "content": "I'm super curious about what it'll take too!
", "rawMarkdown": "I'm super curious about what it'll take too!" } ] }, { "id": 2188152, "postDate": "2023-03-19T10:57:24.623Z", "rawMarkdown": "", "isDeleted": true }, { "id": 2186508, "postDate": "2023-03-17T20:42:54Z", "rawMarkdown": "", "isDeleted": true }, { "id": 2193125, "postDate": "2023-03-23T05:18:49.140Z", "content": "Thanks for sharing!
", "rawMarkdown": "Thanks for sharing!" } ], "comments": [ { "id": 2247315, "author_name": "b ing", "author_url": "", "post_date": "2023-05-05T21:40:56.107000", "content": "Hello, @jpposma, in previous competition at kaggle HuBMAP + HPA - Hacking the Human Body hosts answered to the participant that they are a non-profit organization, so it was one of reasons why SegFormer was admissible with constrained usage licence SegFormer/LICENSE in the competition.
\nSorry, if this is obvious, I don't have enough experience with model licenses for their usage at kaggle yet. Can I use this model in this competition too? Maybe this question will be useful for other people
", "votes": 2, "replies": [ { "id": 2272765, "author_name": "JP Posma", "author_url": "", "post_date": "2023-05-24T18:03:34.570000", "content": "For anyone coming across this, we clarified here (TL;DR: yes you can use those) https://www.kaggle.com/competitions/vesuvius-challenge-ink-detection/discussion/410906
", "votes": 0, "replies": [] } ] }, { "id": 2185362, "author_name": "Halla Yang", "author_url": "", "post_date": "2023-03-17T02:00:50.567000", "content": "This is very cool.
", "votes": 4, "replies": [ { "id": 2185380, "author_name": "Will Cukierski", "author_url": "", "post_date": "2023-03-17T02:36:08.217000", "content": "@hallayang it's been a while since seeing you around the forums/leaderboards. Welcome back! :)
", "votes": 2, "replies": [] } ] }, { "id": 2191587, "author_name": "John Costella", "author_url": "", "post_date": "2023-03-22T04:35:12.423000", "content": "G'day Kaggle gods! Thanks for this super interesting competition!
\nI come asking stupid questions. ๐
\nIt's been 11 years since I've entered a Kaggle comp, and that one got me a job, a job I still have today. (Excluding, I guess, the two later Facebook Recruiting Competitions that I was on the admin team for, which I hope doesn't disqualify me from entering one as a competitor.) (And I hope that having lunch with you guys down at Kaggle HQ near the ball park right after I started at Facebook doesn't disqualify me either.) ๐
\nMy quandary is this: for that 2012 comp I was making up my own ML system. It was in C#, and I had to submit a copy of my code and my documentation and a license for Facebook to use it forever everywhere all at once with my final submission. But before that I could just submit submissions.
\nLong story short, I only just finished building that ML system. (Yeah, not very \"move fast\"; don't get me started.) ๐ Anyway, I was looking to see if there were any new comps to test it out on, and this one looks perfect. And you just kicked it off a week ago.
\nThe problem is that you've made a lot of improvements to Kaggle in the past 11 years. (Duh. What did I expect?) And it looks like it's only possible to make submissions if you do it in a Notebook in Python or R, maybe pulling in some standard AMLT.
\nIs there any way for me to make submissions based on Bear?
\nIf not, that's fine. I only wanted to test out Bear, and even on the off chance that I won anything (which I didn't do last time, although I did get the job, I suppose) I would donate it to some bear-related charity anyway.
\nIn any case, I have the data now (ha, rubbing hands in glee whenever I get a super interesting dataset) ๐ and I'd be happy to report anecdotally whether Bear digests it well or not.
\nThanks in advance!
", "votes": 1, "replies": [ { "id": 2192334, "author_name": "JP Posma", "author_url": "", "post_date": "2023-03-22T15:39:10.417000", "content": "Welcome back to Kaggle! Sure, you can use whatever you want. Good luck!
", "votes": 1, "replies": [ { "id": 2192347, "author_name": "John Costella", "author_url": "", "post_date": "2023-03-22T15:49:40.063000", "content": "But the submissions require use of a Notebook for the code, not just a submission CSV file, right? Or am I misunderstanding the restrictions here?
", "votes": 0, "replies": [] }, { "id": 2192351, "author_name": "JP Posma", "author_url": "", "post_date": "2023-03-22T15:51:25.650000", "content": "Correct. But you can run within that notebook whatever you want.
", "votes": 1, "replies": [] }, { "id": 2192383, "author_name": "John Costella", "author_url": "", "post_date": "2023-03-22T16:13:07.320000", "content": "๐ Ah. Got it. I didn't think that that was legit, but I see that it fits within the comp rules.
\nThanks!
", "votes": 0, "replies": [] }, { "id": 2193013, "author_name": "John Costella", "author_url": "", "post_date": "2023-03-23T03:38:22.730000", "content": "OK, so now that I actually read the Data text properly (see my \"Dumb question about the fragments\" question), I'm confused again.
\nIf the test fragment masks and CT scans are hidden (substituted by mocks in the supplied dataโthat's what I missed when I read it the first few times), then the Notebook has to shell out to Bear itself. But the Notebook is hosted on your end, right? How can I compile a program on that host so that the Notebook can shell out to it and run it?
", "votes": 0, "replies": [] } ] } ] }, { "id": 2183856, "author_name": "hengck23", "author_url": "", "post_date": "2023-03-16T01:03:54", "content": "Hi, thanks for organizing an incredible competition.
\n\" There is a $150,000 grand prize available to the first team that can read these scrolls from a 3d x-ray scan. \"
\nhttps://scrollprize.org/participate
I suppose this is another separate competition. Can this solution uses winning solution from kaggle ink detection?
", "votes": 1, "replies": [ { "id": 2183857, "author_name": "JP Posma", "author_url": "", "post_date": "2023-03-16T01:05:54.650000", "content": "Yes, for that separate competition you can use whatever method that you want. When Ink Detection finishes the winning solutions will be made open source, so you can use those then.
", "votes": 0, "replies": [] } ] }, { "id": 2264336, "author_name": "Jason Heesang Lee", "author_url": "", "post_date": "2023-05-18T10:49:18.933000", "content": "Hi team,
\nI have just started learning ML/DL and all. And this is my first attempt on Kaggle competition, to get a brief understanding of how it goes and how the competitions are like.
\nI am trying to search how should I approach, what model should I use, etc..
\nBut the thing is I have no clue with what keyword I should search with.
\nI tried to search ink detection, image detection, 3D image classification, these didn't work out well.
\nCould anyone help me with understanding the concept and find if there is any stackoverflow queries, youtube videos, or any material where I can learn more about it?
\nThanks
I think this competition is highly meaningful. Although my knowledge and experience may not be sufficient, I would like to see the solutions from everyone else.
", "votes": 0, "replies": [] }, { "id": 2256320, "author_name": "Kostiantyn Maksymov", "author_url": "", "post_date": "2023-05-12T11:42:33.180000", "content": "Hi @danielhavir @jpposma it is very frustrating that we still do not have answers about this topic https://www.kaggle.com/competitions/vesuvius-challenge-ink-detection/discussion/407355- can you give us an answer or at least give some estimate when you will decide what to do with segformer models (for me it look that competition is non profit and we can use any models)
", "votes": 0, "replies": [] }, { "id": 2254110, "author_name": "Jai Chauhan", "author_url": "", "post_date": "2023-05-10T16:55:31.293000", "content": "Anyone here.. Who hires the data scientist in his/her team?
\nI have good knowledge about data science but I have no experience about this industry further I am really excited to work as a data scientist.
\nI am a second year student of computer science and I have a diploma of data science with NLP domain.
\nI do not need money I need some experience.
Great Competition!
", "votes": 0, "replies": [] }, { "id": 2243952, "author_name": "Abhishek Vijayan", "author_url": "", "post_date": "2023-05-03T10:13:11.487000", "content": "Thank you for this exciting challenge and for the example submission notebook.
\nHowever I'm unable to save&commit the notebook while using the generated model weights.
\nApologies for cross-posting - I asked this question here (https://www.kaggle.com/competitions/vesuvius-challenge-ink-detection/discussion/406421 ) but didn't get a response. Thought I'd have better chances of getting response here.
\nThanks in advance
As I could understand, there are some channels / discussion on Discord. How do I connect these channels?
", "votes": 0, "replies": [] }, { "id": 2233965, "author_name": "Kathryn Garwood", "author_url": "", "post_date": "2023-04-24T18:16:43.217000", "content": "This is so cool!
", "votes": 0, "replies": [] }, { "id": 2229518, "author_name": "Aman Anand", "author_url": "", "post_date": "2023-04-21T12:49:55.633000", "content": "It's really interesting, although i haven't submitted any of my work till now , but really looking forward to make this one my very first one to the path of gloryโฆ
", "votes": 0, "replies": [] }, { "id": 2199647, "author_name": "Jiaye Zou", "author_url": "", "post_date": "2023-03-27T22:37:23.243000", "content": "Interesting competition!
", "votes": 0, "replies": [] }, { "id": 2199340, "author_name": "BHOMESH RAZDAN", "author_url": "", "post_date": "2023-03-27T16:37:25.697000", "content": "Advice for beginner in Machine learning
", "votes": 0, "replies": [] }, { "id": 2192641, "author_name": "Farhan Hai Khan", "author_url": "", "post_date": "2023-03-22T19:45:21.603000", "content": "Just some clarification, hosts:
\nI am unable to submit / output ?
", "votes": 0, "replies": [] }, { "id": 2190809, "author_name": "J>Hรธuse Rock", "author_url": "", "post_date": "2023-03-21T14:05:35.373000", "content": "Hey Everyone new to the site / once I submit and run my model and it has completed where can I find status etcโฆ. not seeing myself on leaderboard ?
", "votes": 0, "replies": [ { "id": 2192936, "author_name": "Hao Ji", "author_url": "", "post_date": "2023-03-23T02:03:12.923000", "content": "You can find the submission button on the right-hand side of the Ink Detection competition page. After clicking it, you can see the status of your submission.
\n
\nBy the way, I have a question. I ran and submitted the Vesuvius Challenge: Example submission file in the discussion area, but it keeps showing \"notebook timeout.\" Has anyone else encountered this issue?
\n
I had the same issue. Not sure I have seen anyone share a successful submission of this example. For sure you will need to not train and predict in the same session. So use the provided model or train your own, load it and run just the prediction.
\nI am trying to re-write and re-train using the dual GPU that kaggle provides. That might be fast enough to beat the clock. But getting torch to use two GPU's is often harder than it seems it should be :)
\nSince I have dual GPU on my local machines I am going to follow that path - but you might also play with larger batch sizing to see if you can beat the clock without hitting the memory limit on kaggle.
", "votes": 0, "replies": [] }, { "id": 2193204, "author_name": "Hao Ji", "author_url": "", "post_date": "2023-03-23T06:23:10.780000", "content": "Thanks for your suggestion, I will try it later :)
", "votes": 0, "replies": [] } ] }, { "id": 2195524, "author_name": "Mustafa Tolu", "author_url": "", "post_date": "2023-03-24T17:39:21.243000", "content": "There are some submission rules that you may not obey.
", "votes": 0, "replies": [] } ] }, { "id": 2189924, "author_name": "ilyasse_chaouki", "author_url": "", "post_date": "2023-03-20T22:27:56.083000", "content": "very good !
", "votes": 0, "replies": [] }, { "id": 2189856, "author_name": "Med Amine Chadli", "author_url": "", "post_date": "2023-03-20T20:52:43.747000", "content": "very nice
", "votes": 0, "replies": [] }, { "id": 2189449, "author_name": "Salmane Ait-Ssi", "author_url": "", "post_date": "2023-03-20T14:05:22.267000", "content": "very cool
", "votes": 0, "replies": [] }, { "id": 2185984, "author_name": "Jebastin Nadar", "author_url": "", "post_date": "2023-03-17T13:21:41.430000", "content": "\n\nYou can expect two fragments in the hidden test set, which together are roughly the same size as a single training fragment.
\n
So is the private LB scored on only 1 fragment? My concern is that it may lead to solutions which overfit to this one fragment score higher on the final LB than more generalizable solutions.
", "votes": 0, "replies": [ { "id": 2186151, "author_name": "JP Posma", "author_url": "", "post_date": "2023-03-17T15:48:33.993000", "content": "We only had 4 scanned fragments in total, so we didn't have much of a choice, unfortunately. We would have loved more data too, but this is what we have!
", "votes": 2, "replies": [] } ] }, { "id": 2185907, "author_name": "Iurii Uspangaliev", "author_url": "", "post_date": "2023-03-17T12:01:52.220000", "content": "What an interesting competition! Like it is something mystical to learn about this manuscripts from the past. I will invest my time for sure. Thanks host!
", "votes": 0, "replies": [ { "id": 2186149, "author_name": "JP Posma", "author_url": "", "post_date": "2023-03-17T15:47:55.413000", "content": "Thank you :)
", "votes": 0, "replies": [] } ] }, { "id": 2184423, "author_name": "grahamamen obahiagbon", "author_url": "", "post_date": "2023-03-16T11:25:58.893000", "content": "Please take me along while you loop for Vesuvius coding project. I like to see to the end of all its going to take.
", "votes": 0, "replies": [ { "id": 2184769, "author_name": "JP Posma", "author_url": "", "post_date": "2023-03-16T15:54:10.423000", "content": "I'm super curious about what it'll take too!
", "votes": 0, "replies": [] } ] }, { "id": 2188152, "author_name": "", "author_url": "", "post_date": "2023-03-19T10:57:24.623000", "content": "", "votes": 0, "replies": [] }, { "id": 2186508, "author_name": "", "author_url": "", "post_date": "2023-03-17T20:42:54", "content": "", "votes": 0, "replies": [] }, { "id": 2193125, "author_name": "Ashish Kumar", "author_url": "", "post_date": "2023-03-23T05:18:49.140000", "content": "Thanks for sharing!
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "2182539": "Hey, welcome to the Ink Detection Progress Prize of the Vesuvius Challenge!\n\nYou might find the following resources particularly useful:\n* [Tutorial](https://scrollprize.org/tutorial4)\n* [Notebook](https://www.kaggle.com/jpposma/vesuvius-challenge-ink-detection-tutorial/edit)\n* [Example submission](https://www.kaggle.com/code/danielhavir/vesuvius-challenge-example-submission)\n* Other pages on the main [Vesuvius Challenge website](http://scrollprize.org/)\n\nFeel free to ask any and all questions. We're here to help out. Good luck and have fun!\nโ [Daniel](https://www.kaggle.com/danielhavir) & [JP](https://www.kaggle.com/jpposma)", "2247315": "Hello, @jpposma, in previous competition at kaggle [HuBMAP + HPA - Hacking the Human Body](https://www.kaggle.com/competitions/hubmap-organ-segmentation/overview) hosts answered to the participant that they are a [non-profit organization] (https://www.kaggle.com/competitions/hubmap-organ-segmentation/discussion/356201), so it was one of reasons why SegFormer was admissible with constrained usage licence [SegFormer/LICENSE](https://github.com/NVlabs/SegFormer/blob/master/LICENSE) in the competition.\n\nSorry, if this is obvious, I don't have enough experience with model licenses for their usage at kaggle yet. Can I use this model in this competition too? Maybe this question will be useful for other people", "2185362": "This is very cool.", "2191587": "G'day Kaggle gods! Thanks for this super interesting competition!\n\nI come asking stupid questions. ๐\n\nIt's been 11 years since I've entered a Kaggle comp, and [that one got me a job](https://johncostella.com/bear/history/), a job I still have today. (Excluding, I guess, the two later Facebook Recruiting Competitions that I was on the admin team for, which I hope doesn't disqualify me from entering one as a competitor.) (And I hope that having lunch with you guys down at Kaggle HQ near the ball park right after I started at Facebook doesn't disqualify me either.) ๐\n\nMy quandary is this: for that 2012 comp I was making up my own ML system. It was in C#, and I had to submit a copy of my code and my documentation and a license for Facebook to use it forever everywhere all at once with my final submission. But before that I could just submit submissions.\n\nLong story short, [I only just finished building that ML system](https://johncostella.com/bear/). (Yeah, not very \"move fast\"; don't get me started.) ๐ Anyway, I was looking to see if there were any new comps to test it out on, and this one looks perfect. And you just kicked it off a week ago.\n\nThe problem is that you've made a lot of improvements to Kaggle in the past 11 years. (Duh. What did I expect?) And it looks like it's only possible to make submissions if you do it in a Notebook in Python or R, maybe pulling in some standard AMLT.\n\nIs there any way for me to make submissions based on Bear?\n\nIf not, that's fine. I only wanted to test out Bear, and even on the off chance that I won anything (which I didn't do last time, although I did get the job, I suppose) I would donate it to some bear-related charity anyway.\n\nIn any case, I have the data now (ha, rubbing hands in glee whenever I get a super interesting dataset) ๐ and I'd be happy to report anecdotally whether Bear digests it well or not.\n\nThanks in advance!", "2183856": "Hi, thanks for organizing an incredible competition.\n\n\" There is a $150,000 grand prize available to the first team that can read these scrolls from a 3d x-ray scan. \" \nhttps://scrollprize.org/participate\n\nI suppose this is another separate competition. Can this solution uses winning solution from kaggle ink detection?", "2264336": "Hi team,\nI have just started learning ML/DL and all. And this is my first attempt on Kaggle competition, to get a brief understanding of how it goes and how the competitions are like.\nI am trying to search how should I approach, what model should I use, etc.. \nBut the thing is I have no clue with what keyword I should search with.\nI tried to search ink detection, image detection, 3D image classification, these didn't work out well.\nCould anyone help me with understanding the concept and find if there is any stackoverflow queries, youtube videos, or any material where I can learn more about it?\nThanks ", "2259360": "I think this competition is highly meaningful. Although my knowledge and experience may not be sufficient, I would like to see the solutions from everyone else.", "2256320": "Hi @danielhavir @jpposma it is very frustrating that we still do not have answers about this topic https://www.kaggle.com/competitions/vesuvius-challenge-ink-detection/discussion/407355- can you give us an answer or at least give some estimate when you will decide what to do with segformer models (for me it look that competition is non profit and we can use any models)", "2254110": "Anyone here.. Who hires the data scientist in his/her team?\nI have good knowledge about data science but I have no experience about this industry further I am really excited to work as a data scientist.\nI am a second year student of computer science and I have a diploma of data science with NLP domain.\nI do not need money I need some experience. ", "2248386": "Great Competition!", "2243952": "Thank you for this exciting challenge and for the example submission notebook.\n\nHowever I'm unable to save&commit the notebook while using the generated model weights.\nApologies for cross-posting - I asked this question here (https://www.kaggle.com/competitions/vesuvius-challenge-ink-detection/discussion/406421 ) but didn't get a response. Thought I'd have better chances of getting response here. \nThanks in advance", "2241641": "As I could understand, there are some channels / discussion on Discord. How do I connect these channels?", "2233965": "This is so cool!", "2229518": "It's really interesting, although i haven't submitted any of my work till now , but really looking forward to make this one my very first one to the path of glory... ", "2199647": "Interesting competition!", "2199340": "Advice for beginner in Machine learning\n", "2192641": "Just some clarification, hosts:\n1. The infrared images are provided just for visualisation right, as I see they aren't present in the test folder?\n2. Each of the 65 tifs hold important feature information about the voxel (as I comprehend), disregarding compute/resources as possible constraints, what do these features mean exactly? Are they just the 3D cut-out of each voxel, separately?", "2191337": "I am unable to submit / output ? ", "2190809": "Hey Everyone new to the site / once I submit and run my model and it has completed where can I find status etc.... not seeing myself on leaderboard ? ", "2189924": "very good !", "2189856": "very nice\n", "2189449": "very cool\n", "2185984": "> You can expect two fragments in the hidden test set, which together are roughly the same size as a single training fragment.\n\nSo is the private LB scored on only **1** fragment? My concern is that it may lead to solutions which overfit to this one fragment score higher on the final LB than more generalizable solutions.", "2185907": "What an interesting competition! Like it is something mystical to learn about this manuscripts from the past. I will invest my time for sure. Thanks host!", "2184423": "Please take me along while you loop for Vesuvius coding project. I like to see to the end of all its going to take.", "2188152": "", "2186508": "", "2193125": "Thanks for sharing!" } }