![\"\"](\"https://www.kaggleusercontent.com/kf/123624357/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..vCaeEk7GTgf8_NHEfeb17w.ZiydEpDoXKoS9hZRU8cg5jt6M2XCwT7iusTfCo7pyEqstJwHgtbqIDvx-BAXy_p-X-jVBMb7FMe83tmdLYmQN2N_7NjHECqGH6MyB53kFCn2JdbKFrJnzNFz6wlLeid-VgpgIqJbkOlY5CqD_JrONg9jTEfZoJ-ANddh2mjlvXkOeV4vVtFrZdtrfCISb_v-oaTjLucRucDkkU6yv2jUpW4RVYIhEbeEPI2wz_9bGX89XnJP0ElXirtjNRooq_LEhFnWyUCvJihHNT6F8REGPLO6jL3a_ykDgf3Jp5aOoElOdsm2PP6FzTkTPerrg8sQhgOYy6stg-zX2ZA6FGMYBesuXKY7Kj-iE6KY634cHjNzDcdP5fVcGh7LUfCihEnF5IGAKwZlYrjQvf6q4fY04LjHW0eZkF1G0LRfh5jMdMNSX1St5Ed1ftzbYUEZlTwW85gSj_d9i_xS-Dcf5P57EoxmZ5gQLBxkafRSmhRY_W-sFUjSDpskXTGTZxrxIKmu---T4PisooUgqCpLLDcmgMQIgmL9b6T3FrNrm4_XFjRUhnSvZ9zfmEH6afAg2baGrG-_QsoUUW3ckW8iO3uJ9RYkfkHT-z2rejTODaauZ_QjyYdh6ZyBvkfXdGuk3m7LNu9G4QZ9WtSHjrhG7m0I1J50C5dQjTmmBAtpv1YUO7Vgi6G2QXk0IlrzncQNgyrE.hmavVL1wzj3Nvi3YkjpXMg/__results___files/__results___19_1.png\")
\nThe space in the z-range [30:64] looks like air.
Can someone please confirm my suspicion, that the *.tif files are ordered from deepest in the fragment to surface. E.g. \"00.tif\" contains the voxels farthest from the fragment surface and \"65.tiff\" are the layers on the surface or the air above.
\nLooking at a crosssection (from Efficient image loading with zarr):
\nThe space in the z-range [30:64] looks like air.
Can someone please confirm my suspicion, that the *.tif files are ordered from deepest in the fragment to surface. E.g. \"00.tif\" contains the voxels farthest from the fragment surface and \"65.tiff\" are the layers on the surface or the air above.
\nLooking at a crosssection (from Efficient image loading with zarr):
\nThe space in the z-range [30:64] looks like air.
The surface of the fragment is approximately the middle slice of the stack, so 00 is equally far from the surface as 65. Your basic point is correct, though, that the ordering goes from inside the surface to outside of it.
", "rawMarkdown": "The surface of the fragment is approximately the middle slice of the stack, so 00 is equally far from the surface as 65. Your basic point is correct, though, that the ordering goes from inside the surface to outside of it.", "votes": null }, { "id": "2209312", "postDate": "04/04/2023 15:58:52", "content": "Just to clarify further (apologies if this has already been answered, I couldn't find it in the discussion).
\nWhich side is the ink \"on\"? i.e. in the terminology above, is it on:
\n-- the \"inside\" of the surface, so mostly XX.tif with XX < 30\n-- or on the \"outside\" of the surface, so mostly XX.tif with XX > 30 ?
Thanks in advance.
", "rawMarkdown": "Just to clarify further (apologies if this has already been answered, I couldn't find it in the discussion). \n\nWhich side is the ink \"on\"? i.e. in the terminology above, is it on:\n-- the \"inside\" of the surface, so mostly XX.tif with XX < 30\n-- or on the \"outside\" of the surface, so mostly XX.tif with XX > 30 ?\n\nThanks in advance.", "votes": null }, { "id": "2209324", "postDate": "04/04/2023 16:03:35", "content": "A little of both / it's an open question.
", "rawMarkdown": "A little of both / it's an open question.", "votes": null }, { "id": "2209326", "postDate": "04/04/2023 16:05:52", "content": "Ah sorry, I meant something a bit more mundane -- maybe I really mean: which side was the IR photo taken from?
", "rawMarkdown": "Ah sorry, I meant something a bit more mundane -- maybe I really mean: which side was the IR photo taken from?", "votes": null }, { "id": "2209417", "postDate": "04/04/2023 17:05:55", "content": "The photo is taken from the side containing open air. The dataset slices go from 00 (inside the surface) to 64 (outside the surface, open air), so the photograph was taken (in some vague sense) from the perspective of slice 64.
", "rawMarkdown": "The photo is taken from the side containing open air. The dataset slices go from 00 (inside the surface) to 64 (outside the surface, open air), so the photograph was taken (in some vague sense) from the perspective of slice 64.", "votes": null }, { "id": "2209520", "postDate": "04/04/2023 18:15:12", "content": "Thank you!
", "rawMarkdown": "Thank you!", "votes": null }, { "id": "2224351", "postDate": "04/17/2023 08:30:22", "content": "So, ink sides are the slice 64 for all train dataset, as we can see that the characters e.g., \"P\"s, \"K\"s, or \"E\"s are not flipped.
\nOn the other hand, we are not sure that this assumption is valid for test dataset.
\nIs it true? Or can the host clarify the characters are not flipped for test dataset too?
The test dataset might or might not be flipped or rotated. But the z-axis is the same.
", "rawMarkdown": "The test dataset might or might not be flipped or rotated. But the z-axis is the same.", "votes": null }, { "id": "2225148", "postDate": "04/17/2023 23:55:17", "content": "Thank you for your reply. It must be very natural situation.
\nCan I ask which side is inside the roll? I thought ink side is inside, as I would roll them in this way.
Yes, typically the ink is on the inside of a scroll, though the fragments that we're working with here have detached from their scrolls and have been scanned by themselves.
", "rawMarkdown": "Yes, typically the ink is on the inside of a scroll, though the fragments that we're working with here have detached from their scrolls and have been scanned by themselves.", "votes": null } ], "comments": [ { "id": 2202267, "author_name": "csparker", "author_url": "", "post_date": "03/29/2023 21:32:55", "content": "The surface of the fragment is approximately the middle slice of the stack, so 00 is equally far from the surface as 65. Your basic point is correct, though, that the ordering goes from inside the surface to outside of it.
", "votes": null, "replies": [ { "id": 2209312, "author_name": "nblqbl", "author_url": "", "post_date": "04/04/2023 15:58:52", "content": "Just to clarify further (apologies if this has already been answered, I couldn't find it in the discussion).
\nWhich side is the ink \"on\"? i.e. in the terminology above, is it on:
\n-- the \"inside\" of the surface, so mostly XX.tif with XX < 30\n-- or on the \"outside\" of the surface, so mostly XX.tif with XX > 30 ?
Thanks in advance.
", "votes": null, "replies": [ { "id": 2209324, "author_name": "stephenrparsons", "author_url": "", "post_date": "04/04/2023 16:03:35", "content": "A little of both / it's an open question.
", "votes": null, "replies": [ { "id": 2209326, "author_name": "nblqbl", "author_url": "", "post_date": "04/04/2023 16:05:52", "content": "Ah sorry, I meant something a bit more mundane -- maybe I really mean: which side was the IR photo taken from?
", "votes": null, "replies": [ { "id": 2209417, "author_name": "csparker", "author_url": "", "post_date": "04/04/2023 17:05:55", "content": "The photo is taken from the side containing open air. The dataset slices go from 00 (inside the surface) to 64 (outside the surface, open air), so the photograph was taken (in some vague sense) from the perspective of slice 64.
", "votes": null, "replies": [ { "id": 2209520, "author_name": "nblqbl", "author_url": "", "post_date": "04/04/2023 18:15:12", "content": "Thank you!
", "votes": null, "replies": [ { "id": 2224351, "author_name": "tomooinubushi", "author_url": "", "post_date": "04/17/2023 08:30:22", "content": "So, ink sides are the slice 64 for all train dataset, as we can see that the characters e.g., \"P\"s, \"K\"s, or \"E\"s are not flipped.
\nOn the other hand, we are not sure that this assumption is valid for test dataset.
\nIs it true? Or can the host clarify the characters are not flipped for test dataset too?
The test dataset might or might not be flipped or rotated. But the z-axis is the same.
", "votes": null, "replies": [ { "id": 2225148, "author_name": "tomooinubushi", "author_url": "", "post_date": "04/17/2023 23:55:17", "content": "Thank you for your reply. It must be very natural situation.
\nCan I ask which side is inside the roll? I thought ink side is inside, as I would roll them in this way.
Yes, typically the ink is on the inside of a scroll, though the fragments that we're working with here have detached from their scrolls and have been scanned by themselves.
", "votes": null, "replies": [] } ] } ] } ] } ] } ] } ] } ] } ] } ] } ], "raw_markdown_by_id": { "2202261": "Can someone please confirm my suspicion, that the *.tif files are ordered from deepest in the fragment to surface. E.g. \"00.tif\" contains the voxels farthest from the fragment surface and \"65.tiff\" are the layers on the surface or the air above.\n\nLooking at a crosssection (from [Efficient image loading with zarr](https://www.kaggle.com/code/brettolsen/efficient-image-loading-with-zarr)): \nThe space in the z-range [30:64] looks like air.", "2202267": "The surface of the fragment is approximately the middle slice of the stack, so 00 is equally far from the surface as 65. Your basic point is correct, though, that the ordering goes from inside the surface to outside of it.", "2209312": "Just to clarify further (apologies if this has already been answered, I couldn't find it in the discussion). \n\nWhich side is the ink \"on\"? i.e. in the terminology above, is it on:\n-- the \"inside\" of the surface, so mostly XX.tif with XX < 30\n-- or on the \"outside\" of the surface, so mostly XX.tif with XX > 30 ?\n\nThanks in advance.", "2209324": "A little of both / it's an open question.", "2209326": "Ah sorry, I meant something a bit more mundane -- maybe I really mean: which side was the IR photo taken from?", "2209417": "The photo is taken from the side containing open air. The dataset slices go from 00 (inside the surface) to 64 (outside the surface, open air), so the photograph was taken (in some vague sense) from the perspective of slice 64.", "2209520": "Thank you!", "2224351": "So, ink sides are the slice 64 for all train dataset, as we can see that the characters e.g., \"P\"s, \"K\"s, or \"E\"s are not flipped. \nOn the other hand, we are not sure that this assumption is valid for test dataset. \nIs it true? Or can the host clarify the characters are not flipped for test dataset too?", "2224878": "The test dataset might or might not be flipped or rotated. But the z-axis is the same.", "2225148": "Thank you for your reply. It must be very natural situation.\nCan I ask which side is inside the roll? I thought ink side is inside, as I would roll them in this way.", "2225175": "Yes, typically the ink is on the inside of a scroll, though the fragments that we're working with here have detached from their scrolls and have been scanned by themselves." }, "source": "meta" }