![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F20686390%2Faa44b0545de38f34ddb573e0bf3dcc47%2FScreenshot%202025-03-08%20at%208.25.41AM.png?generation=1741440421195405&alt=media\")
It seems there is an inconsistency in the train_labels.csv file. Specifically, for certain tomo_ids, the \"Number of motors\" is listed as 2, yet there is only one corresponding row. The tomo_ids that seem to have issues are as follows:
\n\ntomo_2b3cdf, \ntomo_62eea8, \ntomo_c84b8e, \ntomo_e6f7f7\n
I have made a code to verify the above. Please check it.
\nhttps://www.kaggle.com/code/chack3/possible-inconsistencies-in-the-train-labels-csv
It seems there is an inconsistency in the train_labels.csv file. Specifically, for certain tomo_ids, the \"Number of motors\" is listed as 2, yet there is only one corresponding row. The tomo_ids that seem to have issues are as follows:
\n\ntomo_2b3cdf, \ntomo_62eea8, \ntomo_c84b8e, \ntomo_e6f7f7\n
I have made a code to verify the above. Please check it.
\nhttps://www.kaggle.com/code/chack3/possible-inconsistencies-in-the-train-labels-csv
To add on top of that, if you inspect these examples visually it appears that there should in fact be two labeled motors, but one of the labels are missing. Here's a screenshot of tomo_e6f7f7, the red point is the existing motor label, but it looks like there should be another label directly to the right.
\n
There was always single flagellum on anatomy pictures that I checked, but I guess there could be any number of them. I think each of them are powered by their own motors, so yeah the other one should be labeled as well.
", "rawMarkdown": "There was always single flagellum on anatomy pictures that I checked, but I guess there could be any number of them. I think each of them are powered by their own motors, so yeah the other one should be labeled as well.", "votes": null }, { "id": "3147673", "postDate": "03/12/2025 09:27:38", "content": "Is there any update regarding those inconsistencies? @jacksonpond @braxtonowens
", "rawMarkdown": "Is there any update regarding those inconsistencies? @jacksonpond @braxtonowens", "votes": null }, { "id": "3147721", "postDate": "03/12/2025 11:01:08", "content": "Thanks for the great finding!
", "rawMarkdown": "Thanks for the great finding!", "votes": null }, { "id": "3147964", "postDate": "03/12/2025 15:48:29", "content": "Thank you, @chack3, for bringing this to our attention. I've discussed the issue with the Kaggle team, and unfortunately, it isn’t an easy fix on our side. I’ll pin this comment so others know about the inconsistency and the manual workaround.
", "rawMarkdown": "Thank you, @chack3, for bringing this to our attention. I've discussed the issue with the Kaggle team, and unfortunately, it isn’t an easy fix on our side. I’ll pin this comment so others know about the inconsistency and the manual workaround.", "votes": null }, { "id": "3154935", "postDate": "03/20/2025 14:35:20", "content": "Couldn't it be on this specific picture that the flagellum is twisted and it's the same one that we see going back and forth ?
", "rawMarkdown": "Couldn't it be on this specific picture that the flagellum is twisted and it's the same one that we see going back and forth ?", "votes": null }, { "id": "3158383", "postDate": "03/24/2025 13:57:41", "content": "My thought also
", "rawMarkdown": "My thought also", "votes": null }, { "id": "3159041", "postDate": "03/25/2025 07:35:56", "content": "Thanks all for the insights and discussion. I noticed this issue while checking consistency between Number of motors and actual label count (e.g., tomo_2b3cdf, tomo_e6f7f7). Given the visual cues suggesting a second unlabeled motor, could this be a case of occlusion or annotation miss during dataset construction?
\nCurious if similar patterns exist in s1K or s1-32B, has anyone checked?
", "rawMarkdown": "Thanks all for the insights and discussion. I noticed this issue while checking consistency between Number of motors and actual label count (e.g., tomo_2b3cdf, tomo_e6f7f7). Given the visual cues suggesting a second unlabeled motor, could this be a case of occlusion or annotation miss during dataset construction?\n\nCurious if similar patterns exist in s1K or s1-32B, has anyone checked?", "votes": null }, { "id": "3159760", "postDate": "03/25/2025 21:47:35", "content": "Tomo_fe050c also seems to have two motors.
", "rawMarkdown": "Tomo_fe050c also seems to have two motors.", "votes": null }, { "id": "3160528", "postDate": "03/26/2025 22:06:28", "content": "Thanks for the great finding and all your work and sharing.
", "rawMarkdown": "Thanks for the great finding and all your work and sharing.", "votes": null }, { "id": "3187771", "postDate": "04/26/2025 13:57:27", "content": "Since the images are the result of cryogenic electron tomography, they are frozen in place. So no motion possible.
", "rawMarkdown": "Since the images are the result of cryogenic electron tomography, they are frozen in place. So no motion possible.", "votes": null } ], "comments": [ { "id": 3144536, "author_name": "johnpayne0", "author_url": "", "post_date": "03/08/2025 13:28:39", "content": "To add on top of that, if you inspect these examples visually it appears that there should in fact be two labeled motors, but one of the labels are missing. Here's a screenshot of tomo_e6f7f7, the red point is the existing motor label, but it looks like there should be another label directly to the right.
\n
There was always single flagellum on anatomy pictures that I checked, but I guess there could be any number of them. I think each of them are powered by their own motors, so yeah the other one should be labeled as well.
", "votes": null, "replies": [ { "id": 3147673, "author_name": "gunesevitan", "author_url": "", "post_date": "03/12/2025 09:27:38", "content": "Is there any update regarding those inconsistencies? @jacksonpond @braxtonowens
", "votes": null, "replies": [ { "id": 3147964, "author_name": "braxtonowens", "author_url": "", "post_date": "03/12/2025 15:48:29", "content": "Thank you, @chack3, for bringing this to our attention. I've discussed the issue with the Kaggle team, and unfortunately, it isn’t an easy fix on our side. I’ll pin this comment so others know about the inconsistency and the manual workaround.
", "votes": null, "replies": [] } ] } ] }, { "id": 3154935, "author_name": "jeanmaxrodriguez", "author_url": "", "post_date": "03/20/2025 14:35:20", "content": "Couldn't it be on this specific picture that the flagellum is twisted and it's the same one that we see going back and forth ?
", "votes": null, "replies": [ { "id": 3158383, "author_name": "edonisraci", "author_url": "", "post_date": "03/24/2025 13:57:41", "content": "My thought also
", "votes": null, "replies": [] }, { "id": 3187771, "author_name": "nickspanhook", "author_url": "", "post_date": "04/26/2025 13:57:27", "content": "Since the images are the result of cryogenic electron tomography, they are frozen in place. So no motion possible.
", "votes": null, "replies": [] } ] } ] }, { "id": 3147721, "author_name": "jahid012", "author_url": "", "post_date": "03/12/2025 11:01:08", "content": "Thanks for the great finding!
", "votes": null, "replies": [] }, { "id": 3159041, "author_name": "achievement", "author_url": "", "post_date": "03/25/2025 07:35:56", "content": "Thanks all for the insights and discussion. I noticed this issue while checking consistency between Number of motors and actual label count (e.g., tomo_2b3cdf, tomo_e6f7f7). Given the visual cues suggesting a second unlabeled motor, could this be a case of occlusion or annotation miss during dataset construction?
\nCurious if similar patterns exist in s1K or s1-32B, has anyone checked?
", "votes": null, "replies": [] }, { "id": 3159760, "author_name": "gabrielferri", "author_url": "", "post_date": "03/25/2025 21:47:35", "content": "Tomo_fe050c also seems to have two motors.
", "votes": null, "replies": [] }, { "id": 3160528, "author_name": "", "author_url": "", "post_date": "03/26/2025 22:06:28", "content": "Thanks for the great finding and all your work and sharing.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "3144398": "braxtonowens\n\nIt seems there is an inconsistency in the train_labels.csv file. Specifically, for certain tomo_ids, the \"Number of motors\" is listed as 2, yet there is only one corresponding row. The tomo_ids that seem to have issues are as follows:\n\n``\ntomo_2b3cdf, \ntomo_62eea8, \ntomo_c84b8e, \ntomo_e6f7f7\n``\n\nI have made a code to verify the above. Please check it.\nhttps://www.kaggle.com/code/chack3/possible-inconsistencies-in-the-train-labels-csv", "3144536": "To add on top of that, if you inspect these examples visually it appears that there should in fact be two labeled motors, but one of the labels are missing. Here's a screenshot of tomo_e6f7f7, the red point is the existing motor label, but it looks like there should be another label directly to the right.\n\n", "3144541": "There was always single flagellum on anatomy pictures that I checked, but I guess there could be any number of them. I think each of them are powered by their own motors, so yeah the other one should be labeled as well.", "3147673": "Is there any update regarding those inconsistencies? @jacksonpond @braxtonowens", "3147721": "Thanks for the great finding!", "3147964": "Thank you, @chack3, for bringing this to our attention. I've discussed the issue with the Kaggle team, and unfortunately, it isn’t an easy fix on our side. I’ll pin this comment so others know about the inconsistency and the manual workaround.", "3154935": "Couldn't it be on this specific picture that the flagellum is twisted and it's the same one that we see going back and forth ?", "3158383": "My thought also", "3159041": "Thanks all for the insights and discussion. I noticed this issue while checking consistency between Number of motors and actual label count (e.g., tomo_2b3cdf, tomo_e6f7f7). Given the visual cues suggesting a second unlabeled motor, could this be a case of occlusion or annotation miss during dataset construction?\n\nCurious if similar patterns exist in s1K or s1-32B, has anyone checked?", "3159760": "Tomo_fe050c also seems to have two motors.", "3160528": "Thanks for the great finding and all your work and sharing.", "3187771": "Since the images are the result of cryogenic electron tomography, they are frozen in place. So no motion possible." }, "source": "meta" }