![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F5055010%2Fc5a5ccd1d41f5517743aa6a4524c31ee%2Fcluster6.png?generation=1600691177768400&alt=media\")
I am sharing here the most informative images from 3D CT scan.
\nMany of us trying to get most out of 3D images. However, there could be good insights in 2D images.
Link to -> 512x512 2D most informative (pixel wise) images
\nHere is result of clustering using EfficientNetB5 and KMeans
\n
There are some odd IDS found.
\nHere are some with no information
\nBAD_IDS = ['ID00026637202179561894768','ID00128637202219474716089','ID00132637202222178761324']\nAnd here are some IDS that are different from majority (don't know they are bad or not)
\nSUSPICIOUS_IDS=['ID00027637202179689871102','ID00078637202199415319443','ID00283637202278714365037']\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F5055010%2F612406b2c9b0b53770eaffe795b84daa%2Fanomalies.png?generation=1600691394891784&alt=media\")
This dataset is created from Carlos Souza's Auto-Encoder notebook
", "messages": [ { "id": 1020792, "postDate": "2020-09-21T12:38:07.923Z", "content": "I am sharing here the most informative images from 3D CT scan.
\nMany of us trying to get most out of 3D images. However, there could be good insights in 2D images.
Link to -> 512x512 2D most informative (pixel wise) images
\nHere is result of clustering using EfficientNetB5 and KMeans
\n
There are some odd IDS found.
\nHere are some with no information
\nBAD_IDS = ['ID00026637202179561894768','ID00128637202219474716089','ID00132637202222178761324']\nAnd here are some IDS that are different from majority (don't know they are bad or not)
\nSUSPICIOUS_IDS=['ID00027637202179689871102','ID00078637202199415319443','ID00283637202278714365037']\n
This dataset is created from Carlos Souza's Auto-Encoder notebook
", "rawMarkdown": "I am sharing here the most informative images from 3D CT scan.\nMany of us trying to get most out of 3D images. However, there could be good insights in 2D images.\n\n[Link to -> 512x512 2D most informative (pixel wise) images](https://www.kaggle.com/bayartsogtya/osic-pulmonary-masked-2d-images)\n\n## Initial insights:\nHere is result of clustering using EfficientNetB5 and KMeans\n\n\n\nThere are some odd IDS found.\n\n## Odd Scans (cluster 1 & cluster 5):\nHere are some with no information\n```\nBAD_IDS = ['ID00026637202179561894768','ID00128637202219474716089','ID00132637202222178761324']\n```\n\nAnd here are some IDS that are different from majority (don't know they are bad or not)\n```\nSUSPICIOUS_IDS=['ID00027637202179689871102','ID00078637202199415319443','ID00283637202278714365037']\n```\n\n\n## Acknowledgement\nThis dataset is created from [Carlos Souza's Auto-Encoder notebook](https://www.kaggle.com/carlossouza/osic-autoencoder-training)", "votes": 6 } ], "comments": [], "raw_markdown_by_id": { "1020792": "I am sharing here the most informative images from 3D CT scan.\nMany of us trying to get most out of 3D images. However, there could be good insights in 2D images.\n\n[Link to -> 512x512 2D most informative (pixel wise) images](https://www.kaggle.com/bayartsogtya/osic-pulmonary-masked-2d-images)\n\n## Initial insights:\nHere is result of clustering using EfficientNetB5 and KMeans\n\n\n\nThere are some odd IDS found.\n\n## Odd Scans (cluster 1 & cluster 5):\nHere are some with no information\n```\nBAD_IDS = ['ID00026637202179561894768','ID00128637202219474716089','ID00132637202222178761324']\n```\n\nAnd here are some IDS that are different from majority (don't know they are bad or not)\n```\nSUSPICIOUS_IDS=['ID00027637202179689871102','ID00078637202199415319443','ID00283637202278714365037']\n```\n\n\n## Acknowledgement\nThis dataset is created from [Carlos Souza's Auto-Encoder notebook](https://www.kaggle.com/carlossouza/osic-autoencoder-training)" } }