![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4472591%2Ff31fcb9d5d7b6aa92159c0d4237ebc42%2F2024-12-30%20225619.png?generation=1735567051626596&alt=media\")
In many tomographies, there are black spots near z=35 that are not labeled, which are about the same size as particles. What are these? Particles of interest?
\n
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4472591%2Fdbfdfcab436dbddd4c44b5a2334ab064%2F2024-12-30%202259531.png?generation=1735567658366641&alt=media\")
In many tomographies, there are black spots near z=35 that are not labeled, which are about the same size as particles. What are these? Particles of interest?
\n
The actual label is a point. Any other volumetric masks are just approximations based on assuming spherical particles of known radius.
", "rawMarkdown": "The actual label is a point. Any other volumetric masks are just approximations based on assuming spherical particles of known radius.", "votes": null }, { "id": "3084456", "postDate": "12/30/2024 20:29:42", "content": "See the discussion here:
\n\n\nIt might be useful to note that the top and bottom z-edges of the tomogram are typically the most affected by artifacts during the imaging process and contain little meaningful information. If I recall correctly (others can confirm), this is because these regions are directly exposed to the environment. It is typical practice to ignore the 10-20 slices at the top and the bottom.
\n
https://www.kaggle.com/competitions/czii-cryo-et-object-identification/discussion/546329
\nAlthough I guess it depends on what you mean by z = 35. Is that layer 3 or layer 35? Although a few of them look a lot like what's in the first layer of TS_6_4 which are probably just artifacts:
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F10704200%2Fe6833ab3c18d6cdbf60f8223f8ffa36c%2Fartifacts.png?generation=1735591001038755&alt=media\")
About that. Recently I've started working with synthetic data, and the volumes are 200x630x630. Don't you think since trainig is 184x630x630 maybe those slices have been already discarted?
", "rawMarkdown": "About that. Recently I've started working with synthetic data, and the volumes are 200x630x630. Don't you think since trainig is 184x630x630 maybe those slices have been already discarted?", "votes": null }, { "id": "3086177", "postDate": "01/02/2025 04:41:49", "content": "Maybe, but the comment seems consistent with the image from the discussion in which the first labeled ribosome shows up in layer 7 low res which is roughly layer 28 in high res. And then the last shows up in layer 35 low res or ~40 layers from the end in high res. Not sure about the other particles though.
", "rawMarkdown": "Maybe, but the comment seems consistent with the image from the discussion in which the first labeled ribosome shows up in layer 7 low res which is roughly layer 28 in high res. And then the last shows up in layer 35 low res or ~40 layers from the end in high res. Not sure about the other particles though.", "votes": null }, { "id": "3086932", "postDate": "01/02/2025 21:26:52", "content": "@sacuscreed If you set the first 10 and last 10 layers of your prediction mask to zero it hurts your score so maybe you are correct.
", "rawMarkdown": "sacuscreed If you set the first 10 and last 10 layers of your prediction mask to zero it hurts your score so maybe you are correct.", "votes": null } ], "comments": [ { "id": 3084199, "author_name": "sacuscreed", "author_url": "", "post_date": "12/30/2024 14:26:33", "content": "The actual label is a point. Any other volumetric masks are just approximations based on assuming spherical particles of known radius.
", "votes": null, "replies": [] }, { "id": 3084456, "author_name": "davidlist", "author_url": "", "post_date": "12/30/2024 20:29:42", "content": "See the discussion here:
\n\n\nIt might be useful to note that the top and bottom z-edges of the tomogram are typically the most affected by artifacts during the imaging process and contain little meaningful information. If I recall correctly (others can confirm), this is because these regions are directly exposed to the environment. It is typical practice to ignore the 10-20 slices at the top and the bottom.
\n
https://www.kaggle.com/competitions/czii-cryo-et-object-identification/discussion/546329
\nAlthough I guess it depends on what you mean by z = 35. Is that layer 3 or layer 35? Although a few of them look a lot like what's in the first layer of TS_6_4 which are probably just artifacts:
\n
About that. Recently I've started working with synthetic data, and the volumes are 200x630x630. Don't you think since trainig is 184x630x630 maybe those slices have been already discarted?
", "votes": null, "replies": [ { "id": 3086177, "author_name": "davidlist", "author_url": "", "post_date": "01/02/2025 04:41:49", "content": "Maybe, but the comment seems consistent with the image from the discussion in which the first labeled ribosome shows up in layer 7 low res which is roughly layer 28 in high res. And then the last shows up in layer 35 low res or ~40 layers from the end in high res. Not sure about the other particles though.
", "votes": null, "replies": [] }, { "id": 3086932, "author_name": "davidlist", "author_url": "", "post_date": "01/02/2025 21:26:52", "content": "@sacuscreed If you set the first 10 and last 10 layers of your prediction mask to zero it hurts your score so maybe you are correct.
", "votes": null, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "3084185": "In many tomographies, there are black spots near z=35 that are not labeled, which are about the same size as particles. What are these? Particles of interest?\n\n\n\n", "3084199": "The actual label is a point. Any other volumetric masks are just approximations based on assuming spherical particles of known radius.", "3084456": "See the discussion here:\n\n>It might be useful to note that the top and bottom z-edges of the tomogram are typically the most affected by artifacts during the imaging process and contain little meaningful information. If I recall correctly (others can confirm), this is because these regions are directly exposed to the environment. It is typical practice to ignore the 10-20 slices at the top and the bottom.\n\n[https://www.kaggle.com/competitions/czii-cryo-et-object-identification/discussion/546329](https://www.kaggle.com/competitions/czii-cryo-et-object-identification/discussion/546329)\n\nAlthough I guess it depends on what you mean by z = 35. Is that layer 3 or layer 35? Although a few of them look a lot like what's in the first layer of TS_6_4 which are probably just artifacts:\n\n", "3085690": "About that. Recently I've started working with synthetic data, and the volumes are 200x630x630. Don't you think since trainig is 184x630x630 maybe those slices have been already discarted?", "3086177": "Maybe, but the comment seems consistent with the image from the discussion in which the first labeled ribosome shows up in layer 7 low res which is roughly layer 28 in high res. And then the last shows up in layer 35 low res or ~40 layers from the end in high res. Not sure about the other particles though.", "3086932": "sacuscreed If you set the first 10 and last 10 layers of your prediction mask to zero it hurts your score so maybe you are correct." }, "source": "meta" }