![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F5444116%2Ff75b89e6025fb44268468a3a8d94dd1f%2F2023-08-01%20225338.png?generation=1690898207850974&alt=media\")
\n
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F5444116%2Fe7d36843d7b2a7fdbcef4810027ef4af%2F2023-08-01%20225354.png?generation=1690898220613863&alt=media\")
In hopes of sparking some insights, I've attempted to display the metadata on the map.
\nHere are the results.
\n
\n
However, it didn't provide any particular insights for me.
\nI'm sharing this notebook in the hope that this attempt might be useful to someone.
https://www.kaggle.com/code/tomokihirose/visualize-metadata-by-convert-to-lat-lon
\n
Update: After receiving your comment, I have corrected the code for converting latitude and longitude.
", "messages": [ { "id": 2359761, "postDate": "2023-07-26T11:11:35.810Z", "content": "In hopes of sparking some insights, I've attempted to display the metadata on the map.
\nHere are the results.
\n
\n
However, it didn't provide any particular insights for me.
\nI'm sharing this notebook in the hope that this attempt might be useful to someone.
https://www.kaggle.com/code/tomokihirose/visualize-metadata-by-convert-to-lat-lon
\n
Update: After receiving your comment, I have corrected the code for converting latitude and longitude.
", "rawMarkdown": "In hopes of sparking some insights, I've attempted to display the metadata on the map.\n\nHere are the results.\n\n\n\n\nHowever, it didn't provide any particular insights for me. \nI'm sharing this notebook in the hope that this attempt might be useful to someone.\n\n[https://www.kaggle.com/code/tomokihirose/visualize-metadata-by-convert-to-lat-lon\n](https://www.kaggle.com/code/tomokihirose/visualize-metadata-by-convert-to-lat-lon)\n\nUpdate: After receiving your comment, I have corrected the code for converting latitude and longitude.", "votes": 21 }, { "id": 2367652, "postDate": "2023-07-31T17:26:10.013Z", "content": "@tomokihirose thanks for sharing. It seems you confused latitude and longitude. It is to be transposed.
\nlat, lon = lon, lat\n\"To avoid large viewing angles, we restrict the
\nsampling of image patches to be between -50 to 50 degree
\nlatitude and -135 to -30 degree longitude.\" - https://arxiv.org/pdf/2304.02122.pdf
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F6259210%2Fa549bae49131a5fc9ee011bf46bb1e58%2F2304.02122.pdf%20-%20Google%20Chrome%202023-07-31%2013.20.30.png?generation=1690824367960898&alt=media\")
Hi, @sergiosaharovskiy
\nThank you for pointing that out.
\nI had overlooked the fact that the angle was being limited.
\n\nIt seems you confused latitude and longitude. It is to be transposed.
\n
When you say it's confusing, are you referring to the variable names in the Notebook?
\nIf that's the case, I've already made the correction but the result hasn't changed. Thank you.
@tomokihirose I mean your current visualization stretches points from Pacific Ocean to Africa, it is to be transposed (latitude is to be swapped with longitude). Here is how it was supposed to look like:![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F6259210%2Fd676c172298c2027244e1354e98834be%2FScreenshot%20from%202023-08-01%2008-41-02.png?generation=1690893891009006&alt=media\")
and not like this:![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F6259210%2Ff933c5e0a205ede7d5a7d9404b1d2dd4%2FScreenshot%20from%202023-08-01%2008-47-50.png?generation=1690894119591478&alt=media\")
Thank you for your detailed explanation.
\nI was looking at the data after converting to latitude and longitude and was confused, but it seems I made a mistake in specifying the order of rows and columns.
Thank you for sharing! I like it a lot!
\nI hadn't realised until now that it is actually a geostationary satellite and it seems to be over Colombia. This may be important because the incidence angle won't be the same for the images taken right below the satellite than for the ones taken far West or East.
\nI wonder if this could be the reason why data augmentations don't seem to work.
Only HFlip works for me
", "rawMarkdown": "Only HFlip works for me", "votes": 1 } ] }, { "id": 2363958, "postDate": "2023-07-29T03:47:50.667Z", "content": "Great work!
", "rawMarkdown": "Great work!" }, { "id": 2361329, "postDate": "2023-07-27T10:45:43.413Z", "content": "Amazing work.
", "rawMarkdown": "Amazing work." } ], "comments": [ { "id": 2367652, "author_name": "SSS", "author_url": "", "post_date": "2023-07-31T17:26:10.013000", "content": "@tomokihirose thanks for sharing. It seems you confused latitude and longitude. It is to be transposed.
\nlat, lon = lon, lat\n\"To avoid large viewing angles, we restrict the
\nsampling of image patches to be between -50 to 50 degree
\nlatitude and -135 to -30 degree longitude.\" - https://arxiv.org/pdf/2304.02122.pdf
\n
Hi, @sergiosaharovskiy
\nThank you for pointing that out.
\nI had overlooked the fact that the angle was being limited.
\n\nIt seems you confused latitude and longitude. It is to be transposed.
\n
When you say it's confusing, are you referring to the variable names in the Notebook?
\nIf that's the case, I've already made the correction but the result hasn't changed. Thank you.
@tomokihirose I mean your current visualization stretches points from Pacific Ocean to Africa, it is to be transposed (latitude is to be swapped with longitude). Here is how it was supposed to look like:
and not like this:
Thank you for your detailed explanation.
\nI was looking at the data after converting to latitude and longitude and was confused, but it seems I made a mistake in specifying the order of rows and columns.
Thank you for sharing! I like it a lot!
\nI hadn't realised until now that it is actually a geostationary satellite and it seems to be over Colombia. This may be important because the incidence angle won't be the same for the images taken right below the satellite than for the ones taken far West or East.
\nI wonder if this could be the reason why data augmentations don't seem to work.
Only HFlip works for me
", "votes": 1, "replies": [] } ] }, { "id": 2363958, "author_name": "mabaoguoMBG", "author_url": "", "post_date": "2023-07-29T03:47:50.667000", "content": "Great work!
", "votes": 0, "replies": [] }, { "id": 2361329, "author_name": "ynhuhu", "author_url": "", "post_date": "2023-07-27T10:45:43.413000", "content": "Amazing work.
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "2359761": "In hopes of sparking some insights, I've attempted to display the metadata on the map.\n\nHere are the results.\n\n\n\n\nHowever, it didn't provide any particular insights for me. \nI'm sharing this notebook in the hope that this attempt might be useful to someone.\n\n[https://www.kaggle.com/code/tomokihirose/visualize-metadata-by-convert-to-lat-lon\n](https://www.kaggle.com/code/tomokihirose/visualize-metadata-by-convert-to-lat-lon)\n\nUpdate: After receiving your comment, I have corrected the code for converting latitude and longitude.", "2367652": "@tomokihirose thanks for sharing. It seems you confused latitude and longitude. It is to be transposed.\n\n```python\n\nlat, lon = lon, lat\n```\n\n\"To avoid large viewing angles, we restrict the\nsampling of image patches to be between -50 to 50 degree\nlatitude and -135 to -30 degree longitude.\" - https://arxiv.org/pdf/2304.02122.pdf\n\n", "2360285": "Thank you for sharing! I like it a lot! \nI hadn't realised until now that it is actually a geostationary satellite and it seems to be over Colombia. This may be important because the incidence angle won't be the same for the images taken right below the satellite than for the ones taken far West or East.\nI wonder if this could be the reason why data augmentations don't seem to work.", "2363958": "Great work!", "2361329": "Amazing work." } }