I see some public notebooks where preprocessing is applied where we are identifying dominant hand and after that manipulating the data based on which hand is the dominant hand.
\nCan someone explain what is being done and how it helps?
\nFor example this public noebook by irohith
\nedit : tagging @irohith
", "messages": [ { "id": 2309649, "postDate": "2023-06-19T19:13:21.130Z", "content": "In my notebook, I am assuming 2 things - fingerspelling is done using only one hand and it is done using the dominant hand. After analyzing a few sequences we can see that the non-dominant hand has more missing frames than the dominant hand. So I am selecting the dominant hand by looking at the no. of non nan frames. Low no. of non nan frames = dominant hand.
", "rawMarkdown": "In my notebook, I am assuming 2 things - fingerspelling is done using only one hand and it is done using the dominant hand. After analyzing a few sequences we can see that the non-dominant hand has more missing frames than the dominant hand. So I am selecting the dominant hand by looking at the no. of non nan frames. Low no. of non nan frames = dominant hand.", "votes": 1 }, { "id": 2308984, "postDate": "2023-06-19T10:40:12.570Z", "content": "I see some public notebooks where preprocessing is applied where we are identifying dominant hand and after that manipulating the data based on which hand is the dominant hand.
\nCan someone explain what is being done and how it helps?
\nFor example this public noebook by irohith
\nedit : tagging @irohith
", "rawMarkdown": "I see some public notebooks where preprocessing is applied where we are identifying dominant hand and after that manipulating the data based on which hand is the dominant hand.\n\nCan someone explain what is being done and how it helps?\n\nFor example [this](https://www.kaggle.com/code/irohith/aslfr-transformer) public noebook by [irohith](https://www.kaggle.com/irohith)\n\nedit : tagging @irohith ", "votes": 1 }, { "id": 2309158, "postDate": "2023-06-19T12:54:15.360Z", "content": "Well, it just as you wrote- he did it this way and it worked. If you think that you have a better way to handle the data, just try it, and we will be very happy if you succeed and publish you solution…
", "rawMarkdown": "Well, it just as you wrote- he did it this way and it worked. If you think that you have a better way to handle the data, just try it, and we will be very happy if you succeed and publish you solution...", "replies": [ { "id": 2309596, "postDate": "2023-06-19T18:12:59.783Z", "content": "I understand that but I want to know the reason behind this and the transformation applied which I couldn't understand.
\n@irohith can you explain this or direct us to some resource where it can be found?
In my notebook, I am assuming 2 things - fingerspelling is done using only one hand and it is done using the dominant hand. After analyzing a few sequences we can see that the non-dominant hand has more missing frames than the dominant hand. So I am selecting the dominant hand by looking at the no. of non nan frames. Low no. of non nan frames = dominant hand.
", "votes": 1, "replies": [] }, { "id": 2309158, "author_name": "greySnow", "author_url": "", "post_date": "2023-06-19T12:54:15.360000", "content": "Well, it just as you wrote- he did it this way and it worked. If you think that you have a better way to handle the data, just try it, and we will be very happy if you succeed and publish you solution…
", "votes": 0, "replies": [ { "id": 2309596, "author_name": "Rohit Chaudhari", "author_url": "", "post_date": "2023-06-19T18:12:59.783000", "content": "I understand that but I want to know the reason behind this and the transformation applied which I couldn't understand.
\n@irohith can you explain this or direct us to some resource where it can be found?