https://drive.google.com/drive/folders/1fRvOtAseTIMaFcGchJsddVI3_W-S6n-8?usp=sharing
\nThese are our results. Results seem pretty good for us However when we submit our work we got 0.000 score on leaderboard.
\nIs there anyone got same result like us?
\nIs there a problem with competition evaluation metric?
\nThank You
https://drive.google.com/drive/folders/1fRvOtAseTIMaFcGchJsddVI3_W-S6n-8?usp=sharing
\nThese are our results. Results seem pretty good for us However when we submit our work we got 0.000 score on leaderboard.
\nIs there anyone got same result like us?
\nIs there a problem with competition evaluation metric?
\nThank You
@firaterdemdogan
\nThe goal of this competition is \"instance segmentation\", not \"semantic segmentation\".
\nSo, we need to predict cell masks separately.
As you can see on the evaluation page, submission.csv has prediction of single cell mask in each row.
", "rawMarkdown": "@firaterdemdogan \nThe goal of this competition is \"instance segmentation\", not \"semantic segmentation\".\nSo, we need to predict cell masks separately.\n\nAs you can see on [the evaluation page](https://www.kaggle.com/c/sartorius-cell-instance-segmentation/overview/evaluation), submission.csv has prediction of single cell mask in each row.", "votes": 7, "replies": [ { "id": 1548685, "postDate": "2021-10-18T12:48:59.280Z", "content": "You are right, thanks.
", "rawMarkdown": "You are right, thanks.", "votes": 1 } ] }, { "id": 1548938, "postDate": "2021-10-18T16:36:05.753Z", "content": "Hey there - As @inoueu1 noted this is the result of some possible confusion with the evaluation metric. We'll work to provide additional clarity on the Evaluation page.
", "rawMarkdown": "Hey there - As @inoueu1 noted this is the result of some possible confusion with the evaluation metric. We'll work to provide additional clarity on the Evaluation page.", "votes": 3 }, { "id": 1549973, "postDate": "2021-10-19T10:29:10.937Z", "content": "I note that @inoueu1 has said here that:
\n\"In submission, we have to encoding mask. Evaluation page show the pixels are numbered from top to bottom, then left to right. But, this is not correct. As you know, the training annotations provided are numbered from left to right, then top to bottom, and this is also the case with submission.
\nAn LB score with encoded mask (top -> bottom, left -> right) was 0.0000, but a score with (left -> right, top -> bottom) was over 0.22.
\nUnfortunately, the evaluation page is just a copy from Data Science Bowl 2018. I hope that the competition hosts will update the evaluation page for this competition.\"
Also note that the submission header on the Evaluation page is wrong. It says:ImageId,EncodedPixels, but it should be id,predicted (as illustrated in the sample_submission.csv).
FYI @addisonhoward
", "rawMarkdown": "Also note that the submission header on the Evaluation page is wrong. It says:`ImageId,EncodedPixels`, but it should be `id,predicted` (as illustrated in the `sample_submission.csv`). \n\nFYI @addisonhoward " }, { "id": 1548063, "postDate": "2021-10-17T23:22:29.260Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 1548666, "author_name": "Inoichan", "author_url": "", "post_date": "2021-10-18T12:28:40.783000", "content": "@firaterdemdogan
\nThe goal of this competition is \"instance segmentation\", not \"semantic segmentation\".
\nSo, we need to predict cell masks separately.
As you can see on the evaluation page, submission.csv has prediction of single cell mask in each row.
", "votes": 7, "replies": [ { "id": 1548685, "author_name": "Firatus", "author_url": "", "post_date": "2021-10-18T12:48:59.280000", "content": "You are right, thanks.
", "votes": 1, "replies": [] } ] }, { "id": 1548938, "author_name": "Addison Howard", "author_url": "", "post_date": "2021-10-18T16:36:05.753000", "content": "Hey there - As @inoueu1 noted this is the result of some possible confusion with the evaluation metric. We'll work to provide additional clarity on the Evaluation page.
", "votes": 3, "replies": [] }, { "id": 1549973, "author_name": "John Mitchell", "author_url": "", "post_date": "2021-10-19T10:29:10.937000", "content": "I note that @inoueu1 has said here that:
\n\"In submission, we have to encoding mask. Evaluation page show the pixels are numbered from top to bottom, then left to right. But, this is not correct. As you know, the training annotations provided are numbered from left to right, then top to bottom, and this is also the case with submission.
\nAn LB score with encoded mask (top -> bottom, left -> right) was 0.0000, but a score with (left -> right, top -> bottom) was over 0.22.
\nUnfortunately, the evaluation page is just a copy from Data Science Bowl 2018. I hope that the competition hosts will update the evaluation page for this competition.\"
Also note that the submission header on the Evaluation page is wrong. It says:ImageId,EncodedPixels, but it should be id,predicted (as illustrated in the sample_submission.csv).
FYI @addisonhoward
", "votes": 0, "replies": [] }, { "id": 1548063, "author_name": "", "author_url": "", "post_date": "2021-10-17T23:22:29.260000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1548031": "https://drive.google.com/drive/folders/1fRvOtAseTIMaFcGchJsddVI3_W-S6n-8?usp=sharing\n\nThese are our results. Results seem pretty good for us However when we submit our work we got 0.000 score on leaderboard.\nIs there anyone got same result like us?\nIs there a problem with competition evaluation metric?\nThank You", "1548666": "@firaterdemdogan \nThe goal of this competition is \"instance segmentation\", not \"semantic segmentation\".\nSo, we need to predict cell masks separately.\n\nAs you can see on [the evaluation page](https://www.kaggle.com/c/sartorius-cell-instance-segmentation/overview/evaluation), submission.csv has prediction of single cell mask in each row.", "1548938": "Hey there - As @inoueu1 noted this is the result of some possible confusion with the evaluation metric. We'll work to provide additional clarity on the Evaluation page.", "1549973": "I note that @inoueu1 has said [here](https://www.kaggle.com/c/sartorius-cell-instance-segmentation/discussion/279790) that:\n\n\"In submission, we have to encoding mask. Evaluation page show the pixels are numbered from top to bottom, then left to right. But, this is not correct. As you know, the training annotations provided are numbered from left to right, then top to bottom, and this is also the case with submission.\nAn LB score with encoded mask (top -> bottom, left -> right) was 0.0000, but a score with (left -> right, top -> bottom) was over 0.22.\nUnfortunately, the evaluation page is just a copy from Data Science Bowl 2018. I hope that the competition hosts will update the evaluation page for this competition.\"", "1549445": "Also note that the submission header on the Evaluation page is wrong. It says:`ImageId,EncodedPixels`, but it should be `id,predicted` (as illustrated in the `sample_submission.csv`). \n\nFYI @addisonhoward ", "1548063": "" } }