Hello
\nWhile trying to include SH-SHY5Y cell line data for training, I tried to read the annotation file for the train set in the LIVECell dataset. However, an error is raised while reading the annotation file. I already sent a GitHub issue (#9) in the LIVECell repository, but it seems a previous issue (#4) reported the same error, they solved the issue by downloading and reading the file a second time using the same method described here. From this solution it seems that the original annotation file was corrupted, but I am not sure if this is the case. Does anyone has been able to read this file without errors? or do you know another way to read the annotation file?
\nfrom pycocotools.coco import COCO\n\nannFile = '../input/sartorius-cell-instance-segmentation/LIVECell_dataset_2021/annotations/LIVECell/livecell_coco_train.json'\n\ncoco = COCO(annFile)\nThe following error is raised:
\nloading annotations into memory...\nDone (t=28.31s)\ncreating index...\n\n---------------------------------------------------------------------------\nTypeError Traceback (most recent call last)\n/tmp/ipykernel_57/539935055.py in <module>\n 12 json.dump(dataset, f)\n 13 \n---> 14 check_ANNT = COCO(ANNT_TRAIN_COPY)\n\n/opt/conda/lib/python3.7/site-packages/pycocotools/coco.py in __init__(self, annotation_file)\n 87 print('Done (t={:0.2f}s)'.format(time.time()- tic))\n 88 self.dataset = dataset\n---> 89 self.createIndex()\n 90 \n 91 def createIndex(self):\n\n/opt/conda/lib/python3.7/site-packages/pycocotools/coco.py in createIndex(self)\n 96 if 'annotations' in self.dataset:\n 97 for ann in self.dataset['annotations']:\n---> 98 imgToAnns[ann['image_id']].append(ann)\n 99 anns[ann['id']] = ann\n 100 \n\nTypeError: string indices must be integers\nHello
\nWhile trying to include SH-SHY5Y cell line data for training, I tried to read the annotation file for the train set in the LIVECell dataset. However, an error is raised while reading the annotation file. I already sent a GitHub issue (#9) in the LIVECell repository, but it seems a previous issue (#4) reported the same error, they solved the issue by downloading and reading the file a second time using the same method described here. From this solution it seems that the original annotation file was corrupted, but I am not sure if this is the case. Does anyone has been able to read this file without errors? or do you know another way to read the annotation file?
\nfrom pycocotools.coco import COCO\n\nannFile = '../input/sartorius-cell-instance-segmentation/LIVECell_dataset_2021/annotations/LIVECell/livecell_coco_train.json'\n\ncoco = COCO(annFile)\nThe following error is raised:
\nloading annotations into memory...\nDone (t=28.31s)\ncreating index...\n\n---------------------------------------------------------------------------\nTypeError Traceback (most recent call last)\n/tmp/ipykernel_57/539935055.py in <module>\n 12 json.dump(dataset, f)\n 13 \n---> 14 check_ANNT = COCO(ANNT_TRAIN_COPY)\n\n/opt/conda/lib/python3.7/site-packages/pycocotools/coco.py in __init__(self, annotation_file)\n 87 print('Done (t={:0.2f}s)'.format(time.time()- tic))\n 88 self.dataset = dataset\n---> 89 self.createIndex()\n 90 \n 91 def createIndex(self):\n\n/opt/conda/lib/python3.7/site-packages/pycocotools/coco.py in createIndex(self)\n 96 if 'annotations' in self.dataset:\n 97 for ann in self.dataset['annotations']:\n---> 98 imgToAnns[ann['image_id']].append(ann)\n 99 anns[ann['id']] = ann\n 100 \n\nTypeError: string indices must be integers\nI read these annotations just as json and then converted and saved in appropriate format manually.
", "rawMarkdown": "I read these annotations just as json and then converted and saved in appropriate format manually.", "votes": null }, { "id": "1594622", "postDate": "11/25/2021 03:06:14", "content": "That problem is discussed here
", "rawMarkdown": "That problem is discussed [here](https://www.kaggle.com/c/sartorius-cell-instance-segmentation/discussion/284694)", "votes": null }, { "id": "1595373", "postDate": "11/25/2021 16:27:28", "content": "Thanks. I will try that too.
", "rawMarkdown": "Thanks. I will try that too.", "votes": null }, { "id": "1595378", "postDate": "11/25/2021 16:31:45", "content": "Thanks, I confirm that is exactly the same issue and will try the solution.
", "rawMarkdown": "Thanks, I confirm that is exactly the same issue and will try the solution.", "votes": null }, { "id": "1603520", "postDate": "12/02/2021 14:53:50", "content": "Do you manually generate LiveCell data after observing the data and merge it with the original data?
", "rawMarkdown": "Do you manually generate LiveCell data after observing the data and merge it with the original data?", "votes": null }, { "id": "1612490", "postDate": "12/09/2021 02:47:03", "content": "There is a solution to read the LIVECell annotation files in the discussion: LIVECELL_dataset_2021 annotationfiles are corrupt. Using those files you could create the LIVECell train and validation sets and then merge them with the other sets. The only thing I found out is that there is a Tensorflow method to concatenate two datasets.
", "rawMarkdown": "There is a solution to read the LIVECell annotation files in the discussion: [LIVECELL_dataset_2021 annotationfiles are corrupt](https://www.kaggle.com/c/sartorius-cell-instance-segmentation/discussion/284694). Using those files you could create the LIVECell train and validation sets and then merge them with the other sets. The only thing I found out is that there is a [Tensorflow method to concatenate two datasets](https://www.tensorflow.org/versions/r2.0/api_docs/python/tf/data/Dataset#concatenate).", "votes": null } ], "comments": [ { "id": 1594344, "author_name": "rednikotin", "author_url": "", "post_date": "11/24/2021 18:05:06", "content": "I read these annotations just as json and then converted and saved in appropriate format manually.
", "votes": null, "replies": [ { "id": 1595373, "author_name": "aramos", "author_url": "", "post_date": "11/25/2021 16:27:28", "content": "Thanks. I will try that too.
", "votes": null, "replies": [] }, { "id": 1603520, "author_name": "zaopolearning", "author_url": "", "post_date": "12/02/2021 14:53:50", "content": "Do you manually generate LiveCell data after observing the data and merge it with the original data?
", "votes": null, "replies": [] }, { "id": 1612490, "author_name": "aramos", "author_url": "", "post_date": "12/09/2021 02:47:03", "content": "There is a solution to read the LIVECell annotation files in the discussion: LIVECELL_dataset_2021 annotationfiles are corrupt. Using those files you could create the LIVECell train and validation sets and then merge them with the other sets. The only thing I found out is that there is a Tensorflow method to concatenate two datasets.
", "votes": null, "replies": [] } ] }, { "id": 1594622, "author_name": "tyaiga", "author_url": "", "post_date": "11/25/2021 03:06:14", "content": "That problem is discussed here
", "votes": null, "replies": [ { "id": 1595378, "author_name": "aramos", "author_url": "", "post_date": "11/25/2021 16:31:45", "content": "Thanks, I confirm that is exactly the same issue and will try the solution.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1594320": "Hello \n\nWhile trying to include SH-SHY5Y cell line data for training, I tried to read the annotation file for the train set in the LIVECell dataset. However, an error is raised while reading the annotation file. I already sent a GitHub issue ([#9](https://github.com/sartorius-research/LIVECell/issues/9)) in the LIVECell repository, but it seems a previous issue ([#4](https://github.com/sartorius-research/LIVECell/issues/4)) reported the same error, they solved the issue by downloading and reading the file a second time using the same method described here. From this solution it seems that the original annotation file was corrupted, but I am not sure if this is the case. Does anyone has been able to read this file without errors? or do you know another way to read the annotation file? \n\n```\nfrom pycocotools.coco import COCO\n\nannFile = '../input/sartorius-cell-instance-segmentation/LIVECell_dataset_2021/annotations/LIVECell/livecell_coco_train.json'\n\ncoco = COCO(annFile)\n```\n\n\nThe following error is raised: \n\n```\nloading annotations into memory...\nDone (t=28.31s)\ncreating index...\n\n---------------------------------------------------------------------------\nTypeError Traceback (most recent call last)\n/tmp/ipykernel_57/539935055.py in