![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F12913052%2F14d50936b854eb501715b6242c06bd95%2FXnip2023-07-16_05-08-38.png?generation=1689455390732703&alt=media\")
I'm having trouble predicting masks with unet++.
\nI could only get all the masks, not separate them to get individual masks
\nSo I had only one value in my coded string, which I think may be the reason for my submission score of 0.
\nI would like to know how you guys predict the separate masks when using unet++?
I'm having trouble predicting masks with unet++.
\nI could only get all the masks, not separate them to get individual masks
\nSo I had only one value in my coded string, which I think may be the reason for my submission score of 0.
\nI would like to know how you guys predict the separate masks when using unet++?
72e40acccadf 512 512 0 1.0 eNqVlFt74iAQhv/SgAaltm61XRpUwLUeYmIOmpPWav7/XYckbe3u1YaYdw7AA8OHaZSQCt6H/Ykz6yqqOpoAGGLqb2sZINYAqAOa2mDbqI3ikLqT+dkwgz8A+LJJO1Pd7DzNXDZsqKb6dtxCyaVwztyshLrwbkXh+7G95tsuWR2mqpDOlk0L2fPYpJS9LZOF7PvMzSUP2Esm73ZMpHIQst8HeR+x5718iNlTIocJG8fy14E9RjIBRIjIWWwCuYfFoNl0s+P2hcaGmwI1NcB1U7V+20O3WEWmgxuhGOhgq1PmRz3hu1KfMzT8rKvGYao7y85MJ0tZn4gNaeqZ0XWUwPVBVKNB9eB
This is what my string looks like
", "rawMarkdown": "`72e40acccadf 512 512 0 1.0 eNqVlFt74iAQhv/SgAaltm61XRpUwLUeYmIOmpPWav7/XYckbe3u1YaYdw7AA8OHaZSQCt6H/Ykz6yqqOpoAGGLqb2sZINYAqAOa2mDbqI3ikLqT+dkwgz8A+LJJO1Pd7DzNXDZsqKb6dtxCyaVwztyshLrwbkXh+7G95tsuWR2mqpDOlk0L2fPYpJS9LZOF7PvMzSUP2Esm73ZMpHIQst8HeR+x5718iNlTIocJG8fy14E9RjIBRIjIWWwCuYfFoNl0s+P2hcaGmwI1NcB1U7V+20O3WEWmgxuhGOhgq1PmRz3hu1KfMzT8rKvGYao7y85MJ0tZn4gNaeqZ0XWUwPVBVKNB9eB `\n\nThis is what my string looks like", "votes": null }, { "id": "2345966", "postDate": "07/15/2023 21:12:03", "content": "display(Image.fromarray(mask.astype(np.uint8)*255))
\nbase64_str = self.__encoder(mask)
How do I split my mask?Thanks a lot.
\n
You can use connected components. This is a notebook that uses it
\nhttps://www.kaggle.com/code/hidngnguyna/baseline-unet-semantic-as-instance-segmentation
Thanks, I will check it.
", "rawMarkdown": "Thanks, I will check it.", "votes": null }, { "id": "2346207", "postDate": "07/16/2023 05:55:06", "content": "@tanjiroll thanks a lot, I have solved my problem.
", "rawMarkdown": "tanjiroll thanks a lot, I have solved my problem.", "votes": null }, { "id": "2346482", "postDate": "07/16/2023 09:36:02", "content": "Using connected components is one way to go. Check the skimage.segmentation submodule for some inspiration https://scikit-image.org/docs/stable/api/skimage.segmentation.html
@yassinealouini thanks, i have solved my problem
", "rawMarkdown": "yassinealouini thanks, i have solved my problem", "votes": null }, { "id": "2351163", "postDate": "07/19/2023 21:34:19", "content": "@watchernight i'm facing same, how did u solved it? always getting zero score
", "rawMarkdown": "watchernight i'm facing same, how did u solved it? always getting zero score", "votes": null }, { "id": "2351594", "postDate": "07/20/2023 08:55:23", "content": "You must set the confidence score
", "rawMarkdown": "You must set the confidence score", "votes": null }, { "id": "2351910", "postDate": "07/20/2023 13:21:32", "content": "which part exactly you mean? confidence score?
\ninstead of
\n\n", "rawMarkdown": "which part exactly you mean? confidence score?\n\ninstead of \n>score = 1.0\nused as \n>score = np.mean(pred[m, :, mask_i > 0])", "votes": null } ], "comments": [ { "id": 2345963, "author_name": "watchernight", "author_url": "", "post_date": "07/15/2023 21:00:57", "content": "score = 1.0
\n
\n used as
\n score = np.mean(pred[m, :, mask_i > 0])
72e40acccadf 512 512 0 1.0 eNqVlFt74iAQhv/SgAaltm61XRpUwLUeYmIOmpPWav7/XYckbe3u1YaYdw7AA8OHaZSQCt6H/Ykz6yqqOpoAGGLqb2sZINYAqAOa2mDbqI3ikLqT+dkwgz8A+LJJO1Pd7DzNXDZsqKb6dtxCyaVwztyshLrwbkXh+7G95tsuWR2mqpDOlk0L2fPYpJS9LZOF7PvMzSUP2Esm73ZMpHIQst8HeR+x5718iNlTIocJG8fy14E9RjIBRIjIWWwCuYfFoNl0s+P2hcaGmwI1NcB1U7V+20O3WEWmgxuhGOhgq1PmRz3hu1KfMzT8rKvGYao7y85MJ0tZn4gNaeqZ0XWUwPVBVKNB9eB
This is what my string looks like
", "votes": null, "replies": [] }, { "id": 2345966, "author_name": "watchernight", "author_url": "", "post_date": "07/15/2023 21:12:03", "content": "display(Image.fromarray(mask.astype(np.uint8)*255))
\nbase64_str = self.__encoder(mask)
How do I split my mask?Thanks a lot.
\n
Using connected components is one way to go. Check the skimage.segmentation submodule for some inspiration https://scikit-image.org/docs/stable/api/skimage.segmentation.html
@yassinealouini thanks, i have solved my problem
", "votes": null, "replies": [] } ] } ] }, { "id": 2346011, "author_name": "tanjiroll", "author_url": "", "post_date": "07/15/2023 23:14:36", "content": "You can use connected components. This is a notebook that uses it
\nhttps://www.kaggle.com/code/hidngnguyna/baseline-unet-semantic-as-instance-segmentation
Thanks, I will check it.
", "votes": null, "replies": [ { "id": 2346207, "author_name": "watchernight", "author_url": "", "post_date": "07/16/2023 05:55:06", "content": "@tanjiroll thanks a lot, I have solved my problem.
", "votes": null, "replies": [ { "id": 2351163, "author_name": "seshurajup", "author_url": "", "post_date": "07/19/2023 21:34:19", "content": "@watchernight i'm facing same, how did u solved it? always getting zero score
", "votes": null, "replies": [ { "id": 2351594, "author_name": "bent1e", "author_url": "", "post_date": "07/20/2023 08:55:23", "content": "You must set the confidence score
", "votes": null, "replies": [ { "id": 2351910, "author_name": "seshurajup", "author_url": "", "post_date": "07/20/2023 13:21:32", "content": "which part exactly you mean? confidence score?
\ninstead of
\n\n", "votes": null, "replies": [] } ] } ] } ] } ] } ] } ], "raw_markdown_by_id": { "2345955": "I'm having trouble predicting masks with unet++.\nI could only get all the masks, not separate them to get individual masks\nSo I had only one value in my coded string, which I think may be the reason for my submission score of 0.\nI would like to know how you guys predict the separate masks when using unet++?", "2345963": "`72e40acccadf 512 512 0 1.0 eNqVlFt74iAQhv/SgAaltm61XRpUwLUeYmIOmpPWav7/XYckbe3u1YaYdw7AA8OHaZSQCt6H/Ykz6yqqOpoAGGLqb2sZINYAqAOa2mDbqI3ikLqT+dkwgz8A+LJJO1Pd7DzNXDZsqKb6dtxCyaVwztyshLrwbkXh+7G95tsuWR2mqpDOlk0L2fPYpJS9LZOF7PvMzSUP2Esm73ZMpHIQst8HeR+x5718iNlTIocJG8fy14E9RjIBRIjIWWwCuYfFoNl0s+P2hcaGmwI1NcB1U7V+20O3WEWmgxuhGOhgq1PmRz3hu1KfMzT8rKvGYao7y85MJ0tZn4gNaeqZ0XWUwPVBVKNB9eB `\n\nThis is what my string looks like", "2345966": "`display(Image.fromarray(mask.astype(np.uint8)*255))`\n`base64_str = self.__encoder(mask)`\n\nHow do I split my mask?Thanks a lot.\n\n", "2346011": "You can use connected components. This is a notebook that uses it\nhttps://www.kaggle.com/code/hidngnguyna/baseline-unet-semantic-as-instance-segmentation", "2346188": "Thanks, I will check it.", "2346207": "tanjiroll thanks a lot, I have solved my problem.", "2346482": "Using connected components is one way to go. Check the `skimage.segmentation` submodule for some inspiration https://scikit-image.org/docs/stable/api/skimage.segmentation.html", "2346672": "yassinealouini thanks, i have solved my problem", "2351163": "watchernight i'm facing same, how did u solved it? always getting zero score", "2351594": "You must set the confidence score", "2351910": "which part exactly you mean? confidence score?\n\ninstead of \n>score = 1.0\nused as \n>score = np.mean(pred[m, :, mask_i > 0])" }, "source": "meta" }score = 1.0
\n
\n used as
\n score = np.mean(pred[m, :, mask_i > 0])