My understanding of the competition is that we should be focusing on the hand/finger positions to identify letters/symbols. But I'm seeing columns in the data involving face and pose landmarks. Furthermore, I'm seeing rows where both the left and right-hand landmark data is nan.
\nSo I have a few questions. If anyone could answer, I'd greatly appreciate it.
\nAre we supposed to incorporate facial changes into our models as if these people are cheating at charades? Are we supposed to build models that can identify letters/symbols by the facial landmarks alone?
And are we to somehow deal with the nan in the hand landmark values as unclean/missing data (this is a good spot to mention this is my first competition) or am I misinterpreting the necessity of the data in those columns?
", "messages": [ { "id": 2258161, "postDate": "2023-05-14T00:22:34.200Z", "content": "You can use any of the data provided - so you can choose if you want to use just the hand data, or use the facial and other landmarks as well.
\nI also see quite a few times when both the right and left hand are nan values, so I think it will be especially tricky (or simply impossible if the data just isn't there) to handle those cases, yes.
", "rawMarkdown": "You can use any of the data provided - so you can choose if you want to use just the hand data, or use the facial and other landmarks as well.\n\nI also see quite a few times when both the right and left hand are nan values, so I think it will be especially tricky (or simply impossible if the data just isn't there) to handle those cases, yes.", "votes": 3 }, { "id": 2258138, "postDate": "2023-05-13T23:35:38.760Z", "content": "My understanding of the competition is that we should be focusing on the hand/finger positions to identify letters/symbols. But I'm seeing columns in the data involving face and pose landmarks. Furthermore, I'm seeing rows where both the left and right-hand landmark data is nan.
\nSo I have a few questions. If anyone could answer, I'd greatly appreciate it.
\nAre we supposed to incorporate facial changes into our models as if these people are cheating at charades? Are we supposed to build models that can identify letters/symbols by the facial landmarks alone?
And are we to somehow deal with the nan in the hand landmark values as unclean/missing data (this is a good spot to mention this is my first competition) or am I misinterpreting the necessity of the data in those columns?
", "rawMarkdown": "My understanding of the competition is that we should be focusing on the hand/finger positions to identify letters/symbols. But I'm seeing columns in the data involving face and pose landmarks. Furthermore, I'm seeing rows where both the left and right-hand landmark data is *nan*. \n\nSo I have a few questions. If anyone could answer, I'd greatly appreciate it. \nAre we supposed to incorporate facial changes into our models as if these people are cheating at charades? Are we supposed to build models that can identify letters/symbols by the facial landmarks alone?\n\nAnd are we to somehow deal with the *nan* in the hand landmark values as unclean/missing data (this is a good spot to mention this is my first competition) or am I misinterpreting the necessity of the data in those columns? ", "votes": 1 } ], "comments": [ { "id": 2258161, "author_name": "chris", "author_url": "", "post_date": "2023-05-14T00:22:34.200000", "content": "You can use any of the data provided - so you can choose if you want to use just the hand data, or use the facial and other landmarks as well.
\nI also see quite a few times when both the right and left hand are nan values, so I think it will be especially tricky (or simply impossible if the data just isn't there) to handle those cases, yes.
", "votes": 3, "replies": [] } ], "raw_markdown_by_id": { "2258161": "You can use any of the data provided - so you can choose if you want to use just the hand data, or use the facial and other landmarks as well.\n\nI also see quite a few times when both the right and left hand are nan values, so I think it will be especially tricky (or simply impossible if the data just isn't there) to handle those cases, yes.", "2258138": "My understanding of the competition is that we should be focusing on the hand/finger positions to identify letters/symbols. But I'm seeing columns in the data involving face and pose landmarks. Furthermore, I'm seeing rows where both the left and right-hand landmark data is *nan*. \n\nSo I have a few questions. If anyone could answer, I'd greatly appreciate it. \nAre we supposed to incorporate facial changes into our models as if these people are cheating at charades? Are we supposed to build models that can identify letters/symbols by the facial landmarks alone?\n\nAnd are we to somehow deal with the *nan* in the hand landmark values as unclean/missing data (this is a good spot to mention this is my first competition) or am I misinterpreting the necessity of the data in those columns? " } }