![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F5069700%2F2aaf0bf7df23a5cc9a0386bdd8ffb26a%2Fwsi-mosaic.png?generation=1698719705856610&alt=media\")
Based on these two continued notebooks:
\nI have created the following large dataset Tiles 🖽 of 🔬 Cancer - 2048px | scale 0.25 (52GB) when each original Whole Slice Images (WSI) was split into tiles and scaled-down to 0.25 ratio. Also, empty tiles (more than 90% background) were skipped.
\n
You can further reduce the black background with simple mapping s shown in
\n", "messages": [ { "id": 2505964, "postDate": "2023-10-31T02:32:28.170Z", "content": "Based on these two continued notebooks:
\nI have created the following large dataset Tiles 🖽 of 🔬 Cancer - 2048px | scale 0.25 (52GB) when each original Whole Slice Images (WSI) was split into tiles and scaled-down to 0.25 ratio. Also, empty tiles (more than 90% background) were skipped.
\n
You can further reduce the black background with simple mapping s shown in
\n", "rawMarkdown": "Based on these two continued notebooks:\n1. [Cancer🔬Subtype: decompose 🖽 large image -> tiles](https://www.kaggle.com/code/jirkaborovec/cancer-subtype-decompose-large-image-tiles)\n2. [Cancer🔬Subtype: decompose 🖽 WSI -> tiles /0.25x](https://www.kaggle.com/code/jirkaborovec/cancer-subtype-decompose-wsi-tiles-0-25x)\n\nI have created the following large **dataset [Tiles 🖽 of 🔬 Cancer - 2048px | scale 0.25](https://www.kaggle.com/datasets/jirkaborovec/tiles-of-cancer-2048px-scale-0-25) (52GB)** when each original Whole Slice Images (WSI) was split into tiles and scaled-down to 0.25 ratio. Also, empty tiles (more than 90% background) were skipped.\n\n\n\nYou can further reduce the black background with simple mapping s shown in\n- [Cancer🔬Subtype: tiles🖽 w/ Lightning⚡TIMM models](https://www.kaggle.com/code/jirkaborovec/cancer-subtype-tiles-w-lightning-timm-models)\n- [Show Tiles 🖽 of cancer🔬Subtype](https://www.kaggle.com/code/jirkaborovec/show-tiles-of-cancer-subtype)", "votes": 26 }, { "id": 2515716, "postDate": "2023-11-07T06:49:46.867Z", "content": "thank you for this.
\nso you tiled with 2048 x 2048 and downsized to 512 x 512?
\nis that what scaling .25 means?
that is correct
", "rawMarkdown": "that is correct", "votes": 2, "replies": [ { "id": 2516911, "postDate": "2023-11-08T04:26:16.950Z", "content": "one more question :-)
\nthis further reduction you talk about when you say:
\n\nYou can further reduce the black background with simple mapping s shown in…
\n
I couldn't spot it in respective notebooks, what is it?
", "rawMarkdown": "one more question :-)\nthis **further reduction** you talk about when you say: \n>You can further reduce the black background with simple mapping s shown in...\n\nI couldn't spot it in respective notebooks, what is it?\n\n" }, { "id": 2516972, "postDate": "2023-11-08T05:29:42.540Z", "content": "It is in this notebook, see lines:
\nimg = plt.imread(img_path)\nmask = np.(img[..., :], axis=) == \nimg[mask, :] = \nok, I got it thanks.
\nnr 1. notebook in the original message (late versions) had it already implemented, so I was thinking there was another further magic :-)
@jirkaborovec Are you planning to update your data due to the master data update?
", "rawMarkdown": "@jirkaborovec Are you planning to update your data due to the master data update?", "votes": 1, "replies": [ { "id": 2513236, "postDate": "2023-11-05T10:03:23.380Z", "content": "Not sure what you mean, could you please give me some more context?
\nUpdate, found it and will prepare a patch
", "rawMarkdown": "Not sure what you mean, could you please give me some more context?\n\nUpdate, [found it](https://www.kaggle.com/competitions/UBC-OCEAN/discussion/451892) and will prepare a patch ", "votes": 1, "replies": [ { "id": 2513824, "postDate": "2023-11-05T18:47:24.177Z", "content": "I used your code to process the updated images again. Data: https://www.kaggle.com/datasets/hubert101/ubctiles
", "rawMarkdown": "I used your code to process the updated images again. Data: [https://www.kaggle.com/datasets/hubert101/ubctiles](url)" }, { "id": 2513957, "postDate": "2023-11-05T22:07:04.420Z", "content": "Dataset has been updated
", "rawMarkdown": "Dataset has been updated ", "votes": 2 } ] } ] }, { "id": 2506830, "postDate": "2023-10-31T15:09:13.103Z", "content": "I faced the problem of memory out for error while submitting notebook?. I am trying to take patch of each image ,but how many patches should i take in each image of testing data , Please suggest me ?
\nI have try with 20 patch of each image in testing data but it will giving me an error ? Please give me solution for this ?
I have a successful submission with a 0.34+ score with these notebooks:
\nHi friend, i based in my notebook by install pyvips, but after run second time my code found error \"OSError: Cannot find pyvips.so library\". ¿Maybe do you found same error?
", "rawMarkdown": "Hi friend, i based in my notebook by install pyvips, but after run second time my code found error \"OSError: Cannot find pyvips.so library\". ¿Maybe do you found same error?" }, { "id": 2511578, "postDate": "2023-11-03T18:48:27.420Z", "content": "you many need to use CPU dataset or GPU dataset
", "rawMarkdown": "you many need to use [CPU dataset](https://www.kaggle.com/datasets/jirkaborovec/pyvips-python-and-deb-package) or [GPU dataset](https://www.kaggle.com/datasets/jirkaborovec/pyvips-python-and-deb-package-gpu)", "votes": 2 } ] } ] }, { "id": 2531564, "postDate": "2023-11-20T10:17:17.007Z", "content": "Hello, I am a newbie in this field.
\nIs the label given to the split tile from WSI the same label that WSI has?
\nIf so, wouldn't the split tiles be assigned a label even though they don't have the characteristics of that label?
\nOr would the tiles split from the WSI not be used as labeled training data?
I think you should check this updated dataset with tumor segmentatuon, UBC-OCEAN: Tiles🖽 w/ masks🔬 2048px | scale 0.25
", "rawMarkdown": "I think you should check this updated dataset with tumor segmentatuon, [UBC-OCEAN: Tiles🖽 w/ masks🔬 2048px | scale 0.25](https://www.kaggle.com/datasets/jirkaborovec/ubc-ocean-tiles-w-masks-2048px-scale-0-25 )", "replies": [ { "id": 2531778, "postDate": "2023-11-20T14:19:13.487Z", "content": "Thanks for the reply!
\nI didn't realize the supplemental data had been added.
\nI understood that the tiles split from the WSI are labeled according to the supplemental data, and all tiles cut from the TMA are labeled with the original image label.
Also this explanation to the labeling, https://www.kaggle.com/competitions/UBC-OCEAN/discussion/455890#2529071
", "rawMarkdown": "Also this explanation to the labeling, https://www.kaggle.com/competitions/UBC-OCEAN/discussion/455890#2529071" } ] }, { "id": 2515848, "postDate": "2023-11-07T09:26:28.833Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 2515716, "author_name": "GUNER", "author_url": "", "post_date": "2023-11-07T06:49:46.867000", "content": "thank you for this.
\nso you tiled with 2048 x 2048 and downsized to 512 x 512?
\nis that what scaling .25 means?
that is correct
", "votes": 2, "replies": [ { "id": 2516911, "author_name": "GUNER", "author_url": "", "post_date": "2023-11-08T04:26:16.950000", "content": "one more question :-)
\nthis further reduction you talk about when you say:
\n\nYou can further reduce the black background with simple mapping s shown in…
\n
I couldn't spot it in respective notebooks, what is it?
", "votes": 0, "replies": [] }, { "id": 2516972, "author_name": "Jirka", "author_url": "", "post_date": "2023-11-08T05:29:42.540000", "content": "It is in this notebook, see lines:
\nimg = plt.imread(img_path)\nmask = np.(img[..., :], axis=) == \nimg[mask, :] = \nok, I got it thanks.
\nnr 1. notebook in the original message (late versions) had it already implemented, so I was thinking there was another further magic :-)
@jirkaborovec Are you planning to update your data due to the master data update?
", "votes": 1, "replies": [ { "id": 2513236, "author_name": "Jirka", "author_url": "", "post_date": "2023-11-05T10:03:23.380000", "content": "Not sure what you mean, could you please give me some more context?
\nUpdate, found it and will prepare a patch
", "votes": 1, "replies": [ { "id": 2513824, "author_name": "HuBERT", "author_url": "", "post_date": "2023-11-05T18:47:24.177000", "content": "I used your code to process the updated images again. Data: https://www.kaggle.com/datasets/hubert101/ubctiles
", "votes": 0, "replies": [] }, { "id": 2513957, "author_name": "Jirka", "author_url": "", "post_date": "2023-11-05T22:07:04.420000", "content": "Dataset has been updated
", "votes": 2, "replies": [] } ] } ] }, { "id": 2506830, "author_name": "sunil thite", "author_url": "", "post_date": "2023-10-31T15:09:13.103000", "content": "I faced the problem of memory out for error while submitting notebook?. I am trying to take patch of each image ,but how many patches should i take in each image of testing data , Please suggest me ?
\nI have try with 20 patch of each image in testing data but it will giving me an error ? Please give me solution for this ?
I have a successful submission with a 0.34+ score with these notebooks:
\nHi friend, i based in my notebook by install pyvips, but after run second time my code found error \"OSError: Cannot find pyvips.so library\". ¿Maybe do you found same error?
", "votes": 0, "replies": [] }, { "id": 2511578, "author_name": "Jirka", "author_url": "", "post_date": "2023-11-03T18:48:27.420000", "content": "you many need to use CPU dataset or GPU dataset
", "votes": 2, "replies": [] } ] } ] }, { "id": 2531564, "author_name": "Iforine", "author_url": "", "post_date": "2023-11-20T10:17:17.007000", "content": "Hello, I am a newbie in this field.
\nIs the label given to the split tile from WSI the same label that WSI has?
\nIf so, wouldn't the split tiles be assigned a label even though they don't have the characteristics of that label?
\nOr would the tiles split from the WSI not be used as labeled training data?
I think you should check this updated dataset with tumor segmentatuon, UBC-OCEAN: Tiles🖽 w/ masks🔬 2048px | scale 0.25
", "votes": 0, "replies": [ { "id": 2531778, "author_name": "Iforine", "author_url": "", "post_date": "2023-11-20T14:19:13.487000", "content": "Thanks for the reply!
\nI didn't realize the supplemental data had been added.
\nI understood that the tiles split from the WSI are labeled according to the supplemental data, and all tiles cut from the TMA are labeled with the original image label.
Also this explanation to the labeling, https://www.kaggle.com/competitions/UBC-OCEAN/discussion/455890#2529071
", "votes": 0, "replies": [] } ] }, { "id": 2515848, "author_name": "", "author_url": "", "post_date": "2023-11-07T09:26:28.833000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "2505964": "Based on these two continued notebooks:\n1. [Cancer🔬Subtype: decompose 🖽 large image -> tiles](https://www.kaggle.com/code/jirkaborovec/cancer-subtype-decompose-large-image-tiles)\n2. [Cancer🔬Subtype: decompose 🖽 WSI -> tiles /0.25x](https://www.kaggle.com/code/jirkaborovec/cancer-subtype-decompose-wsi-tiles-0-25x)\n\nI have created the following large **dataset [Tiles 🖽 of 🔬 Cancer - 2048px | scale 0.25](https://www.kaggle.com/datasets/jirkaborovec/tiles-of-cancer-2048px-scale-0-25) (52GB)** when each original Whole Slice Images (WSI) was split into tiles and scaled-down to 0.25 ratio. Also, empty tiles (more than 90% background) were skipped.\n\n\n\nYou can further reduce the black background with simple mapping s shown in\n- [Cancer🔬Subtype: tiles🖽 w/ Lightning⚡TIMM models](https://www.kaggle.com/code/jirkaborovec/cancer-subtype-tiles-w-lightning-timm-models)\n- [Show Tiles 🖽 of cancer🔬Subtype](https://www.kaggle.com/code/jirkaborovec/show-tiles-of-cancer-subtype)", "2515716": "thank you for this.\nso you tiled with 2048 x 2048 and downsized to 512 x 512?\nis that what scaling .25 means?", "2513234": "@jirkaborovec Are you planning to update your data due to the master data update?", "2506830": "I faced the problem of memory out for error while submitting notebook?. I am trying to take patch of each image ,but how many patches should i take in each image of testing data , Please suggest me ?\nI have try with 20 patch of each image in testing data but it will giving me an error ? Please give me solution for this ?", "2531564": "Hello, I am a newbie in this field.\nIs the label given to the split tile from WSI the same label that WSI has?\nIf so, wouldn't the split tiles be assigned a label even though they don't have the characteristics of that label?\nOr would the tiles split from the WSI not be used as labeled training data?", "2515848": "" } }