Hi,
\nI'm struggling with the submission file, specially for creating the \"prediction string'
\nBelow, few lines from my dataframe \"prepare_to_submission\"
id width weight label confidence\n0 9b1d4b27-6946-4b86-a818-8e91029c3dfa 1728 1728 0.0 0.3484996557235718\n1 5af9bb81-719a-4374-9a12-5e5665329df6 1728 1728 0.0 0.9666309952735901\n2 8adb17b7-ce2c-4721-bf34-c1806c72b4d4 2048 2048 4.0 0.19799888134002686\n3 c6379d83-6b05-4a29-8f56-05a381d94d2a 3072 3072 0.0 0.2289401888847351\n4 277b3f6d-099b-4b6d-8592-a06ad6f52beb 2048 2048 0.0 0.0970616340637207\nSo, I've the ID, width, weight, label, and confidence
\nWhen I try to encode I get this string :
\nb'OQAAAGIAAAAxAAAAZAAAADQAAABiAAAAMgAAADcAAAAtAAAANgAAADkAAA......'\nwhich doesn't look pretty much with the one in the sample_submission
\nWhat I've tried :
\nfor c in np_sub[:,[0,1,2,4,5]][:,[0,3,4]]: ## Where np_sub is == prepare_to_submission.values\n print(np.ascontiguousarray(c))\n base64_str = base64.b64encode(np.ascontiguousarray(c))\n print(base64_str)\nMy question is simple, can I get the prediction_string from this dataframe or do I need to use the \"pycocotools\" and mask as mutils?
\nIf you have any ideas, you're welcome, any help is truly appreciated
Hello,
\nThe prediction string for each test image is expected to be in the format:
\n\nLabelA1 ConfidenceA1 EncodedMaskA1 LabelA2 ConfidenceA2 EncodedMaskA2
\n
where an EncodedMask is the encoded binary mask of one cell. You can extract locations of single cells in an image using the cell segmentation tool. The evaluation page has more details about how the binary mask of a cell is expected to be encoded and also some code that may help. Darek's notebook where he extracts the encoded masks for each cell in the test images may also be useful. You could also just use this dataset, but this would score 0 in the private test set.
", "rawMarkdown": "Hello,\nThe prediction string for each test image is expected to be in the format:\n> LabelA1 ConfidenceA1 EncodedMaskA1 LabelA2 ConfidenceA2 EncodedMaskA2\n\nwhere an EncodedMask is the encoded binary mask of one cell. You can extract locations of single cells in an image using the [cell segmentation tool](https://github.com/CellProfiling/HPA-Cell-Segmentation). The [evaluation page](https://www.kaggle.com/c/hpa-single-cell-image-classification/overview/evaluation) has more details about how the binary mask of a cell is expected to be encoded and also some code that may help.[ Darek's notebook](https://www.kaggle.com/thedrcat/hpa-cell-tiles-test-with-enc/notebook) where he extracts the encoded masks for each cell in the test images may also be useful. You could also just use this dataset, but this would score 0 in the private test set.", "votes": 1 }, { "id": 1235721, "postDate": "2021-03-12T12:10:14.873Z", "content": "Hi,
\nI'm struggling with the submission file, specially for creating the \"prediction string'
\nBelow, few lines from my dataframe \"prepare_to_submission\"
id width weight label confidence\n0 9b1d4b27-6946-4b86-a818-8e91029c3dfa 1728 1728 0.0 0.3484996557235718\n1 5af9bb81-719a-4374-9a12-5e5665329df6 1728 1728 0.0 0.9666309952735901\n2 8adb17b7-ce2c-4721-bf34-c1806c72b4d4 2048 2048 4.0 0.19799888134002686\n3 c6379d83-6b05-4a29-8f56-05a381d94d2a 3072 3072 0.0 0.2289401888847351\n4 277b3f6d-099b-4b6d-8592-a06ad6f52beb 2048 2048 0.0 0.0970616340637207\nSo, I've the ID, width, weight, label, and confidence
\nWhen I try to encode I get this string :
\nb'OQAAAGIAAAAxAAAAZAAAADQAAABiAAAAMgAAADcAAAAtAAAANgAAADkAAA......'\nwhich doesn't look pretty much with the one in the sample_submission
\nWhat I've tried :
\nfor c in np_sub[:,[0,1,2,4,5]][:,[0,3,4]]: ## Where np_sub is == prepare_to_submission.values\n print(np.ascontiguousarray(c))\n base64_str = base64.b64encode(np.ascontiguousarray(c))\n print(base64_str)\nMy question is simple, can I get the prediction_string from this dataframe or do I need to use the \"pycocotools\" and mask as mutils?
\nIf you have any ideas, you're welcome, any help is truly appreciated
Hello,
\nThe prediction string for each test image is expected to be in the format:
\n\nLabelA1 ConfidenceA1 EncodedMaskA1 LabelA2 ConfidenceA2 EncodedMaskA2
\n
where an EncodedMask is the encoded binary mask of one cell. You can extract locations of single cells in an image using the cell segmentation tool. The evaluation page has more details about how the binary mask of a cell is expected to be encoded and also some code that may help. Darek's notebook where he extracts the encoded masks for each cell in the test images may also be useful. You could also just use this dataset, but this would score 0 in the private test set.
", "votes": 1, "replies": [] } ], "raw_markdown_by_id": { "1235802": "Hello,\nThe prediction string for each test image is expected to be in the format:\n> LabelA1 ConfidenceA1 EncodedMaskA1 LabelA2 ConfidenceA2 EncodedMaskA2\n\nwhere an EncodedMask is the encoded binary mask of one cell. You can extract locations of single cells in an image using the [cell segmentation tool](https://github.com/CellProfiling/HPA-Cell-Segmentation). The [evaluation page](https://www.kaggle.com/c/hpa-single-cell-image-classification/overview/evaluation) has more details about how the binary mask of a cell is expected to be encoded and also some code that may help.[ Darek's notebook](https://www.kaggle.com/thedrcat/hpa-cell-tiles-test-with-enc/notebook) where he extracts the encoded masks for each cell in the test images may also be useful. You could also just use this dataset, but this would score 0 in the private test set.", "1235721": "Hi, \nI'm struggling with the submission file, specially for creating the \"prediction string'\nBelow, few lines from my dataframe \"prepare_to_submission\" \n```\nid\twidth\tweight\tlabel\tconfidence\n0\t9b1d4b27-6946-4b86-a818-8e91029c3dfa\t1728\t1728\t0.0\t0.3484996557235718\n1\t5af9bb81-719a-4374-9a12-5e5665329df6\t1728\t1728\t0.0\t0.9666309952735901\n2\t8adb17b7-ce2c-4721-bf34-c1806c72b4d4\t2048\t2048\t4.0\t0.19799888134002686\n3\tc6379d83-6b05-4a29-8f56-05a381d94d2a\t3072\t3072\t0.0\t0.2289401888847351\n4\t277b3f6d-099b-4b6d-8592-a06ad6f52beb\t2048\t2048\t0.0\t0.0970616340637207\n```\nSo, I've the ID, width, weight, label, and confidence\n\nWhen I try to encode I get this string :\n\n```\nb'OQAAAGIAAAAxAAAAZAAAADQAAABiAAAAMgAAADcAAAAtAAAANgAAADkAAA......'\n\n```\nwhich doesn't look pretty much with the one in the sample_submission\n\nWhat I've tried :\n\n```\nfor c in np_sub[:,[0,1,2,4,5]][:,[0,3,4]]: ## Where np_sub is == prepare_to_submission.values\n print(np.ascontiguousarray(c))\n base64_str = base64.b64encode(np.ascontiguousarray(c))\n print(base64_str)\n```\n\nMy question is simple, can I get the prediction_string from this dataframe or do I need to use the \"pycocotools\" and mask as mutils?\nIf you have any ideas, you're welcome, any help is truly appreciated\n\n" } }