First, I really appreciate the effort put into creating this competition!
\nI noticed that annotations are missing for lab_id \"MABe22_keypoints\" and \"MABe22_movies\". I was wondering if there's a specific reason for this?
\nAlso, it might be helpful for other participants if this information could be added to the dataset description, as currently we need to search through the discussion threads to find this out. Just a small suggestion that might save others some time!
\n(Apologies if this has already been explained in another discussion thread and I missed it.)
Thanks again for your work on this competition!
", "messages": [ { "id": "3300713", "postDate": "10/11/2025 08:05:24", "content": "\nFirst, I really appreciate the effort put into creating this competition!
\nI noticed that annotations are missing for lab_id \"MABe22_keypoints\" and \"MABe22_movies\". I was wondering if there's a specific reason for this?
\nAlso, it might be helpful for other participants if this information could be added to the dataset description, as currently we need to search through the discussion threads to find this out. Just a small suggestion that might save others some time!
\n(Apologies if this has already been explained in another discussion thread and I missed it.)
Thanks again for your work on this competition!
", "rawMarkdown": "annkennedy05 \n\nFirst, I really appreciate the effort put into creating this competition!\n\nI noticed that annotations are missing for lab_id \"MABe22_keypoints\" and \"MABe22_movies\". I was wondering if there's a specific reason for this?\nAlso, it might be helpful for other participants if this information could be added to the dataset description, as currently we need to search through the discussion threads to find this out. Just a small suggestion that might save others some time!\n(Apologies if this has already been explained in another discussion thread and I missed it.)\n\nThanks again for your work on this competition!", "votes": null }, { "id": "3302145", "postDate": "10/15/2025 06:55:17", "content": "There is only a small subset of original MABe22 dataset with hand annotations. You can refer to original paper to create annotations using heuristics. See this discussion: Are some files corrupt?
", "rawMarkdown": "There is only a small subset of original MABe22 dataset with hand annotations. You can refer to original paper to create annotations using heuristics. See this discussion: [Are some files corrupt?](https://www.kaggle.com/competitions/MABe-mouse-behavior-detection/discussion/608413)", "votes": null }, { "id": "3302229", "postDate": "10/15/2025 11:32:19", "content": "So to confirm. Basically 8k of 8.8k provided \"videos\" are not labeled.
", "rawMarkdown": "So to confirm. Basically 8k of 8.8k provided \"videos\" are not labeled.", "votes": null }, { "id": "3302285", "postDate": "10/15/2025 13:18:54", "content": "Indeed, although they still provide valuable information about the dynamics of how mice move. If you look into the winning solution from our competition in 2021, you'll see that unannotated pose data was very useful for unsupervised feature learning prior to classification.
", "rawMarkdown": "Indeed, although they still provide valuable information about the dynamics of how mice move. If you look into the winning solution from our competition in 2021, you'll see that unannotated pose data was very useful for unsupervised feature learning prior to classification.", "votes": null }, { "id": "3303126", "postDate": "10/17/2025 07:50:54", "content": "Can you please name the 2021 competition for our reference!!
", "rawMarkdown": "Can you please name the 2021 competition for our reference!!", "votes": null }, { "id": "3303221", "postDate": "10/17/2025 12:28:40", "content": "The 2021 competition is the source of the CalMS21 dataset. Our paper on the dataset and winning solutions can be found here: https://arxiv.org/abs/2104.02710
", "rawMarkdown": "The 2021 competition is the source of the CalMS21 dataset. Our paper on the dataset and winning solutions can be found here: https://arxiv.org/abs/2104.02710", "votes": null } ], "comments": [ { "id": 3302145, "author_name": "vasiliyshangichev", "author_url": "", "post_date": "10/15/2025 06:55:17", "content": "There is only a small subset of original MABe22 dataset with hand annotations. You can refer to original paper to create annotations using heuristics. See this discussion: Are some files corrupt?
", "votes": null, "replies": [] }, { "id": 3302229, "author_name": "sacuscreed", "author_url": "", "post_date": "10/15/2025 11:32:19", "content": "So to confirm. Basically 8k of 8.8k provided \"videos\" are not labeled.
", "votes": null, "replies": [ { "id": 3302285, "author_name": "annkennedy05", "author_url": "", "post_date": "10/15/2025 13:18:54", "content": "Indeed, although they still provide valuable information about the dynamics of how mice move. If you look into the winning solution from our competition in 2021, you'll see that unannotated pose data was very useful for unsupervised feature learning prior to classification.
", "votes": null, "replies": [ { "id": 3303126, "author_name": "subbusrikantam", "author_url": "", "post_date": "10/17/2025 07:50:54", "content": "Can you please name the 2021 competition for our reference!!
", "votes": null, "replies": [ { "id": 3303221, "author_name": "annkennedy05", "author_url": "", "post_date": "10/17/2025 12:28:40", "content": "The 2021 competition is the source of the CalMS21 dataset. Our paper on the dataset and winning solutions can be found here: https://arxiv.org/abs/2104.02710
", "votes": null, "replies": [] } ] } ] } ] } ], "raw_markdown_by_id": { "3300713": "annkennedy05 \n\nFirst, I really appreciate the effort put into creating this competition!\n\nI noticed that annotations are missing for lab_id \"MABe22_keypoints\" and \"MABe22_movies\". I was wondering if there's a specific reason for this?\nAlso, it might be helpful for other participants if this information could be added to the dataset description, as currently we need to search through the discussion threads to find this out. Just a small suggestion that might save others some time!\n(Apologies if this has already been explained in another discussion thread and I missed it.)\n\nThanks again for your work on this competition!", "3302145": "There is only a small subset of original MABe22 dataset with hand annotations. You can refer to original paper to create annotations using heuristics. See this discussion: [Are some files corrupt?](https://www.kaggle.com/competitions/MABe-mouse-behavior-detection/discussion/608413)", "3302229": "So to confirm. Basically 8k of 8.8k provided \"videos\" are not labeled.", "3302285": "Indeed, although they still provide valuable information about the dynamics of how mice move. If you look into the winning solution from our competition in 2021, you'll see that unannotated pose data was very useful for unsupervised feature learning prior to classification.", "3303126": "Can you please name the 2021 competition for our reference!!", "3303221": "The 2021 competition is the source of the CalMS21 dataset. Our paper on the dataset and winning solutions can be found here: https://arxiv.org/abs/2104.02710" }, "source": "meta" }