{ "id": 410846, "title": "If you have a good Kaggle algorithm, win $40,000 in 3 easy steps! 😁", "url": "/competitions/vesuvius-challenge-ink-detection/discussion/410846", "author_name": "", "post_date": "2023-05-16T18:59:00.480202Z", "votes": 10, "comment_count": 20, "views": 0, "content": "

Got a good algorithm in this competition? Be sure to try this:

\n
    \n
  1. Run your algorithm on one of the many segments in the real scrolls, e.g. http://dl.ash2txt.org/full-scrolls/Scroll1.volpkg/paths/20230509182749/layers/ (sign up on scrollprize.org to get the login details). Do keep in mind that the resolution here is 8µm instead of 4µm, so you might have to retrain or somehow adjust your algorithm.
    2. \n
  2. Can you find some letters? Minimum 10 letters in a 4 cm^2 area (one pixel = 8µm, so roughly a 7000x7000 pixel area).
    3. \n
  3. If you find some letters, submit for the First Letters prize. The first to make an eligible submission wins $40k, the second wins $10k! Profit!
    4. \n
", "messages": [ { "id": "2262156", "postDate": "05/16/2023 18:59:00", "content": "

Got a good algorithm in this competition? Be sure to try this:

\n
    \n
  1. Run your algorithm on one of the many segments in the real scrolls, e.g. http://dl.ash2txt.org/full-scrolls/Scroll1.volpkg/paths/20230509182749/layers/ (sign up on scrollprize.org to get the login details). Do keep in mind that the resolution here is 8µm instead of 4µm, so you might have to retrain or somehow adjust your algorithm.
    2. \n
  2. Can you find some letters? Minimum 10 letters in a 4 cm^2 area (one pixel = 8µm, so roughly a 7000x7000 pixel area).
    3. \n
  3. If you find some letters, submit for the First Letters prize. The first to make an eligible submission wins $40k, the second wins $10k! Profit!
    4. \n
", "rawMarkdown": "Got a good algorithm in this competition? Be sure to try this:\n1. Run your algorithm on one of the many segments in the real scrolls, e.g. http://dl.ash2txt.org/full-scrolls/Scroll1.volpkg/paths/20230509182749/layers/ (sign up on scrollprize.org to get the login details). Do keep in mind that the resolution here is 8µm instead of 4µm, so you might have to retrain or somehow adjust your algorithm.\n2. Can you find some letters? Minimum 10 letters in a 4 cm^2 area (one pixel = 8µm, so roughly a 7000x7000 pixel area).\n3. If you find some letters, submit for the [First Letters prize](https://scrollprize.org/first_letters). The first to make an eligible submission wins $40k, the second wins $10k! Profit!", "votes": null }, { "id": "2262337", "postDate": "05/16/2023 20:50:20", "content": "

thanks for sharing that :)

", "rawMarkdown": "thanks for sharing that :)", "votes": null }, { "id": "2262549", "postDate": "05/17/2023 02:27:26", "content": "

Could you please explain more about Login? Confuse with the user-password in http://dl.ash2txt.org/full-scrolls/Scroll1.volpkg/paths/20230509182749/layers/

", "rawMarkdown": "Could you please explain more about Login? Confuse with the user-password in http://dl.ash2txt.org/full-scrolls/Scroll1.volpkg/paths/20230509182749/layers/", "votes": null }, { "id": "2262718", "postDate": "05/17/2023 05:33:45", "content": "

You have to register here: https://docs.google.com/forms/d/e/1FAIpQLSf2lCOCwnO1xo0bc1QdlL0a034Uoe7zyjYBY2k33ZHslHE38Q/viewform

", "rawMarkdown": "You have to register here: https://docs.google.com/forms/d/e/1FAIpQLSf2lCOCwnO1xo0bc1QdlL0a034Uoe7zyjYBY2k33ZHslHE38Q/viewform", "votes": null }, { "id": "2263564", "postDate": "05/17/2023 17:27:29", "content": "

I just wanted to ask how long it normally takes to get the login, I've tried getting access a week ago and again this morning (10h ago) and I got the reply that my form was accepted both times but no login details later. Sorry if I'm just overlooking something obvious.

", "rawMarkdown": "I just wanted to ask how long it normally takes to get the login, I've tried getting access a week ago and again this morning (10h ago) and I got the reply that my form was accepted both times but no login details later. Sorry if I'm just overlooking something obvious.", "votes": null }, { "id": "2263867", "postDate": "05/18/2023 01:36:48", "content": "

The login is available immediately in the instructions after submitting the form

", "rawMarkdown": "The login is available immediately in the instructions after submitting the form", "votes": null }, { "id": "2263900", "postDate": "05/18/2023 02:40:32", "content": "

can you share the signup link

", "rawMarkdown": "can you share the signup link", "votes": null }, { "id": "2264194", "postDate": "05/18/2023 08:32:43", "content": "

@jpposma

\n
\n

Do keep in mind that the resolution here is 8µm instead of 4µm

\n
\n

Does this mean only resolution of x, y axis are different, and resolution of z-axis is same?

", "rawMarkdown": "jpposma \n\n> Do keep in mind that the resolution here is 8µm instead of 4µm\n\nDoes this mean only resolution of x, y axis are different, and resolution of z-axis is same?", "votes": null }, { "id": "2269257", "postDate": "05/22/2023 10:38:44", "content": "

anyone has good results or feedback? (i.e. did your kaggle data trained model generalised will to unseen dataset?)

\n", "rawMarkdown": "anyone has good results or feedback? (i.e. did your kaggle data trained model generalised will to unseen dataset?)\n\n- my results are quite ambiguous (i wonder if there is a bug or data domain problem)", "votes": null }, { "id": "2269259", "postDate": "05/22/2023 10:40:07", "content": "

i suppose we can use these new data set for training? (e.g. unsupervised learning or set post processing threshold) ???

", "rawMarkdown": "i suppose we can use these new data set for training? (e.g. unsupervised learning or set post processing threshold) ???", "votes": null }, { "id": "2270048", "postDate": "05/23/2023 00:10:35", "content": "

They said we can’t use other Herculaneum scroll data, so perhaps the answer is no…?
\nHowever, I believe using these data might push the original purpose of host to read unseen scroll to step further.

\n

https://www.kaggle.com/competitions/vesuvius-challenge-ink-detection/discussion/407972#2251983

", "rawMarkdown": "They said we can’t use other Herculaneum scroll data, so perhaps the answer is no…?\nHowever, I believe using these data might push the original purpose of host to read unseen scroll to step further. \n\nhttps://www.kaggle.com/competitions/vesuvius-challenge-ink-detection/discussion/407972#2251983", "votes": null }, { "id": "2270369", "postDate": "05/23/2023 06:31:53", "content": "

@jpposma How does the hosts handles the front and back of the scroll in the Grand Prize scrolls? I believe the model's generalization performance strongly depends on which side is the surface with ink residue.

", "rawMarkdown": "jpposma How does the hosts handles the front and back of the scroll in the Grand Prize scrolls? I believe the model's generalization performance strongly depends on which side is the surface with ink residue.", "votes": null }, { "id": "2272743", "postDate": "05/24/2023 17:44:53", "content": "

No, the resolution is different on all 3 axes

", "rawMarkdown": "No, the resolution is different on all 3 axes", "votes": null }, { "id": "2272747", "postDate": "05/24/2023 17:46:08", "content": "

All data on http://dl.ash2txt.org/ can be used for training.

", "rawMarkdown": "All data on http://dl.ash2txt.org/ can be used for training.", "votes": null }, { "id": "2272749", "postDate": "05/24/2023 17:46:50", "content": "

We don't care what you do with the front and the back of the scroll. All we want is for you to figure out how to read the text, somehow. :)

", "rawMarkdown": "We don't care what you do with the front and the back of the scroll. All we want is for you to figure out how to read the text, somehow. :)", "votes": null }, { "id": "2273002", "postDate": "05/25/2023 00:20:57", "content": "

This means all the axis (x, y, z) has the same resolution?

\n", "rawMarkdown": "This means all the axis (x, y, z) has the same resolution?\n- Grand Prize dataset (8μm)\n- Kaggle dataset (4μm)", "votes": null }, { "id": "2273012", "postDate": "05/25/2023 00:33:07", "content": "

Exactly, yep.

", "rawMarkdown": "Exactly, yep.", "votes": null }, { "id": "2273014", "postDate": "05/25/2023 00:36:59", "content": "

Thank you for replying.

", "rawMarkdown": "Thank you for replying.", "votes": null }, { "id": "2288521", "postDate": "06/05/2023 12:42:32", "content": "

That's cool!

", "rawMarkdown": "That's cool!", "votes": null }, { "id": "2295656", "postDate": "06/11/2023 07:24:20", "content": "

That's so nice!
\nI'll do my best🔥

", "rawMarkdown": "That's so nice!\nI'll do my best🔥", "votes": null }, { "id": "2300424", "postDate": "06/13/2023 07:38:09", "content": "

That's so nice!

", "rawMarkdown": "That's so nice!", "votes": null } ], "comments": [ { "id": 2262337, "author_name": "beubeu4l", "author_url": "", "post_date": "05/16/2023 20:50:20", "content": "

thanks for sharing that :)

", "votes": null, "replies": [] }, { "id": 2262549, "author_name": "dwchen", "author_url": "", "post_date": "05/17/2023 02:27:26", "content": "

Could you please explain more about Login? Confuse with the user-password in http://dl.ash2txt.org/full-scrolls/Scroll1.volpkg/paths/20230509182749/layers/

", "votes": null, "replies": [ { "id": 2262718, "author_name": "jpposma", "author_url": "", "post_date": "05/17/2023 05:33:45", "content": "

You have to register here: https://docs.google.com/forms/d/e/1FAIpQLSf2lCOCwnO1xo0bc1QdlL0a034Uoe7zyjYBY2k33ZHslHE38Q/viewform

", "votes": null, "replies": [] } ] }, { "id": 2263564, "author_name": "raki21", "author_url": "", "post_date": "05/17/2023 17:27:29", "content": "

I just wanted to ask how long it normally takes to get the login, I've tried getting access a week ago and again this morning (10h ago) and I got the reply that my form was accepted both times but no login details later. Sorry if I'm just overlooking something obvious.

", "votes": null, "replies": [ { "id": 2263867, "author_name": "jpposma", "author_url": "", "post_date": "05/18/2023 01:36:48", "content": "

The login is available immediately in the instructions after submitting the form

", "votes": null, "replies": [] } ] }, { "id": 2263900, "author_name": "arunodhayan", "author_url": "", "post_date": "05/18/2023 02:40:32", "content": "

can you share the signup link

", "votes": null, "replies": [] }, { "id": 2264194, "author_name": "tatamikenn", "author_url": "", "post_date": "05/18/2023 08:32:43", "content": "

@jpposma

\n
\n

Do keep in mind that the resolution here is 8µm instead of 4µm

\n
\n

Does this mean only resolution of x, y axis are different, and resolution of z-axis is same?

", "votes": null, "replies": [ { "id": 2272743, "author_name": "jpposma", "author_url": "", "post_date": "05/24/2023 17:44:53", "content": "

No, the resolution is different on all 3 axes

", "votes": null, "replies": [ { "id": 2273002, "author_name": "tatamikenn", "author_url": "", "post_date": "05/25/2023 00:20:57", "content": "

This means all the axis (x, y, z) has the same resolution?

\n", "votes": null, "replies": [ { "id": 2273012, "author_name": "jpposma", "author_url": "", "post_date": "05/25/2023 00:33:07", "content": "

Exactly, yep.

", "votes": null, "replies": [ { "id": 2273014, "author_name": "tatamikenn", "author_url": "", "post_date": "05/25/2023 00:36:59", "content": "

Thank you for replying.

", "votes": null, "replies": [] } ] } ] } ] } ] }, { "id": 2269257, "author_name": "hengck23", "author_url": "", "post_date": "05/22/2023 10:38:44", "content": "

anyone has good results or feedback? (i.e. did your kaggle data trained model generalised will to unseen dataset?)

\n", "votes": null, "replies": [] }, { "id": 2269259, "author_name": "hengck23", "author_url": "", "post_date": "05/22/2023 10:40:07", "content": "

i suppose we can use these new data set for training? (e.g. unsupervised learning or set post processing threshold) ???

", "votes": null, "replies": [ { "id": 2270048, "author_name": "tatamikenn", "author_url": "", "post_date": "05/23/2023 00:10:35", "content": "

They said we can’t use other Herculaneum scroll data, so perhaps the answer is no…?
\nHowever, I believe using these data might push the original purpose of host to read unseen scroll to step further.

\n

https://www.kaggle.com/competitions/vesuvius-challenge-ink-detection/discussion/407972#2251983

", "votes": null, "replies": [] }, { "id": 2272747, "author_name": "jpposma", "author_url": "", "post_date": "05/24/2023 17:46:08", "content": "

All data on http://dl.ash2txt.org/ can be used for training.

", "votes": null, "replies": [] } ] }, { "id": 2270369, "author_name": "tatamikenn", "author_url": "", "post_date": "05/23/2023 06:31:53", "content": "

@jpposma How does the hosts handles the front and back of the scroll in the Grand Prize scrolls? I believe the model's generalization performance strongly depends on which side is the surface with ink residue.

", "votes": null, "replies": [ { "id": 2272749, "author_name": "jpposma", "author_url": "", "post_date": "05/24/2023 17:46:50", "content": "

We don't care what you do with the front and the back of the scroll. All we want is for you to figure out how to read the text, somehow. :)

", "votes": null, "replies": [] } ] }, { "id": 2288521, "author_name": "alexgodlewski", "author_url": "", "post_date": "06/05/2023 12:42:32", "content": "

That's cool!

", "votes": null, "replies": [] }, { "id": 2295656, "author_name": "yuma81", "author_url": "", "post_date": "06/11/2023 07:24:20", "content": "

That's so nice!
\nI'll do my best🔥

", "votes": null, "replies": [] }, { "id": 2300424, "author_name": "pedrovergaraurquijo", "author_url": "", "post_date": "06/13/2023 07:38:09", "content": "

That's so nice!

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2262156": "Got a good algorithm in this competition? Be sure to try this:\n1. Run your algorithm on one of the many segments in the real scrolls, e.g. http://dl.ash2txt.org/full-scrolls/Scroll1.volpkg/paths/20230509182749/layers/ (sign up on scrollprize.org to get the login details). Do keep in mind that the resolution here is 8µm instead of 4µm, so you might have to retrain or somehow adjust your algorithm.\n2. Can you find some letters? Minimum 10 letters in a 4 cm^2 area (one pixel = 8µm, so roughly a 7000x7000 pixel area).\n3. If you find some letters, submit for the [First Letters prize](https://scrollprize.org/first_letters). The first to make an eligible submission wins $40k, the second wins $10k! Profit!", "2262337": "thanks for sharing that :)", "2262549": "Could you please explain more about Login? Confuse with the user-password in http://dl.ash2txt.org/full-scrolls/Scroll1.volpkg/paths/20230509182749/layers/", "2262718": "You have to register here: https://docs.google.com/forms/d/e/1FAIpQLSf2lCOCwnO1xo0bc1QdlL0a034Uoe7zyjYBY2k33ZHslHE38Q/viewform", "2263564": "I just wanted to ask how long it normally takes to get the login, I've tried getting access a week ago and again this morning (10h ago) and I got the reply that my form was accepted both times but no login details later. Sorry if I'm just overlooking something obvious.", "2263867": "The login is available immediately in the instructions after submitting the form", "2263900": "can you share the signup link", "2264194": "jpposma \n\n> Do keep in mind that the resolution here is 8µm instead of 4µm\n\nDoes this mean only resolution of x, y axis are different, and resolution of z-axis is same?", "2269257": "anyone has good results or feedback? (i.e. did your kaggle data trained model generalised will to unseen dataset?)\n\n- my results are quite ambiguous (i wonder if there is a bug or data domain problem)", "2269259": "i suppose we can use these new data set for training? (e.g. unsupervised learning or set post processing threshold) ???", "2270048": "They said we can’t use other Herculaneum scroll data, so perhaps the answer is no…?\nHowever, I believe using these data might push the original purpose of host to read unseen scroll to step further. \n\nhttps://www.kaggle.com/competitions/vesuvius-challenge-ink-detection/discussion/407972#2251983", "2270369": "jpposma How does the hosts handles the front and back of the scroll in the Grand Prize scrolls? I believe the model's generalization performance strongly depends on which side is the surface with ink residue.", "2272743": "No, the resolution is different on all 3 axes", "2272747": "All data on http://dl.ash2txt.org/ can be used for training.", "2272749": "We don't care what you do with the front and the back of the scroll. All we want is for you to figure out how to read the text, somehow. :)", "2273002": "This means all the axis (x, y, z) has the same resolution?\n- Grand Prize dataset (8μm)\n- Kaggle dataset (4μm)", "2273012": "Exactly, yep.", "2273014": "Thank you for replying.", "2288521": "That's cool!", "2295656": "That's so nice!\nI'll do my best🔥", "2300424": "That's so nice!" }, "source": "meta" }