I would like to ask about \"The green filter should hence be used to predict the label, and the other filters are used as references.\" at the Data Description.
\nAbove subject, I have two question:
\nThere is a one-to-one correspondence between the green image and the label, we can classify by only green image, Is this understanding correct?
About \"references\" in the above sentence, does this word mean that it can be segmented using the red, blue and yellow image?
\nOr, Does this word mean that you can better express the characteristics of the label by using these images? Or, both?
I would appreciate it if you could let me know.
", "messages": [ { "id": 1240564, "postDate": "2021-03-16T13:44:26.357Z", "content": "With references, we mean that we use those other channels to give context to the green channel. Without the blue channel for example, we would have no way of knowing where the nuclei are in the image unless the protein in green happened to overlap with the nuclei.
\nSo the label that you are predicting should be the green channel but the other channels are used to give information about the cells so that the information from the green channel means something.
", "rawMarkdown": "With references, we mean that we use those other channels to give context to the green channel. Without the blue channel for example, we would have no way of knowing where the nuclei are in the image unless the protein in green happened to overlap with the nuclei.\n\nSo the label that you are predicting should be the green channel but the other channels are used to give information about the cells so that the information from the green channel means something.", "votes": 1, "replies": [ { "id": 1241131, "postDate": "2021-03-16T23:07:45.050Z", "content": "Thank you for kindly answering the question.
\nIs there any problem with the following understanding?
\none HPA image contains information from various labels in different region.
\nSo, before we predict the label, we need to specific the region of each labels.
\nThat is, we identify the region from references and predict the label from green channel within region.
The references always target the same structures. Blue is Nucleus, Red is Microtubules, and Yellow is Endoplasmic Reticulum, so you don't technically have to identify those regions. Each cell will have all of these.
\nBut yes, you use those channels to understand what you see in the green and by doing so you can identify the green channel label.
Forgive me if I am misunderstanding your question.
", "rawMarkdown": "The references always target the same structures. Blue is Nucleus, Red is Microtubules, and Yellow is Endoplasmic Reticulum, so you don't technically have to identify those regions. Each cell will have all of these. \nBut yes, you use those channels to understand what you see in the green and by doing so you can identify the green channel label.\n\nForgive me if I am misunderstanding your question.", "votes": 1 }, { "id": 1244516, "postDate": "2021-03-19T03:18:10.693Z", "content": "Thank you for your reply. Perhaps I was misunderstanding.
\nThe advice of \"so you don't technically have to identify those regions. Each cell will have all of these. \" resolved the misunderstanding and I understood the task well.
\nKagglers segment only the cell region, and attache the multiple labels in each cell.
Thank you for your polite explanation. I'll do my best.
", "rawMarkdown": "Thank you for your reply. Perhaps I was misunderstanding.\n\nThe advice of \"so you don't technically have to identify those regions. Each cell will have all of these. \" resolved the misunderstanding and I understood the task well.\nKagglers segment only the cell region, and attache the multiple labels in each cell.\n\nThank you for your polite explanation. I'll do my best." } ] }, { "id": 1239957, "postDate": "2021-03-16T06:49:32.860Z", "content": "I would like to ask about \"The green filter should hence be used to predict the label, and the other filters are used as references.\" at the Data Description.
\nAbove subject, I have two question:
\nThere is a one-to-one correspondence between the green image and the label, we can classify by only green image, Is this understanding correct?
About \"references\" in the above sentence, does this word mean that it can be segmented using the red, blue and yellow image?
\nOr, Does this word mean that you can better express the characteristics of the label by using these images? Or, both?
I would appreciate it if you could let me know.
", "rawMarkdown": "I would like to ask about \"The green filter should hence be used to predict the label, and the other filters are used as references.\" at the Data Description.\n\nAbove subject, I have two question:\n\n1. There is a one-to-one correspondence between the green image and the label, we can classify by only green image, Is this understanding correct?\n\n2. About \"references\" in the above sentence, does this word mean that it can be segmented using the red, blue and yellow image?\nOr, Does this word mean that you can better express the characteristics of the label by using these images? Or, both?\n\nI would appreciate it if you could let me know.", "votes": 2 } ], "comments": [ { "id": 1240564, "author_name": "Casper Winsnes", "author_url": "", "post_date": "2021-03-16T13:44:26.357000", "content": "With references, we mean that we use those other channels to give context to the green channel. Without the blue channel for example, we would have no way of knowing where the nuclei are in the image unless the protein in green happened to overlap with the nuclei.
\nSo the label that you are predicting should be the green channel but the other channels are used to give information about the cells so that the information from the green channel means something.
", "votes": 1, "replies": [ { "id": 1241131, "author_name": "furu-nag", "author_url": "", "post_date": "2021-03-16T23:07:45.050000", "content": "Thank you for kindly answering the question.
\nIs there any problem with the following understanding?
\none HPA image contains information from various labels in different region.
\nSo, before we predict the label, we need to specific the region of each labels.
\nThat is, we identify the region from references and predict the label from green channel within region.
The references always target the same structures. Blue is Nucleus, Red is Microtubules, and Yellow is Endoplasmic Reticulum, so you don't technically have to identify those regions. Each cell will have all of these.
\nBut yes, you use those channels to understand what you see in the green and by doing so you can identify the green channel label.
Forgive me if I am misunderstanding your question.
", "votes": 1, "replies": [] }, { "id": 1244516, "author_name": "furu-nag", "author_url": "", "post_date": "2021-03-19T03:18:10.693000", "content": "Thank you for your reply. Perhaps I was misunderstanding.
\nThe advice of \"so you don't technically have to identify those regions. Each cell will have all of these. \" resolved the misunderstanding and I understood the task well.
\nKagglers segment only the cell region, and attache the multiple labels in each cell.
Thank you for your polite explanation. I'll do my best.
", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "1240564": "With references, we mean that we use those other channels to give context to the green channel. Without the blue channel for example, we would have no way of knowing where the nuclei are in the image unless the protein in green happened to overlap with the nuclei.\n\nSo the label that you are predicting should be the green channel but the other channels are used to give information about the cells so that the information from the green channel means something.", "1239957": "I would like to ask about \"The green filter should hence be used to predict the label, and the other filters are used as references.\" at the Data Description.\n\nAbove subject, I have two question:\n\n1. There is a one-to-one correspondence between the green image and the label, we can classify by only green image, Is this understanding correct?\n\n2. About \"references\" in the above sentence, does this word mean that it can be segmented using the red, blue and yellow image?\nOr, Does this word mean that you can better express the characteristics of the label by using these images? Or, both?\n\nI would appreciate it if you could let me know." } }