{ "id": 109051, "title": "Simple Descriptions of Cloud Types / Labeling Process", "url": "/competitions/understanding_cloud_organization/discussion/109051", "author_name": "", "post_date": "2019-09-16T03:59:51.220464Z", "votes": 31, "comment_count": 14, "views": 0, "content": "

Hi friends!\nJust in cases that somebody like me still gets confused on the definitions of cloud types.\nE.g. I didn't understand different between Sugar vs. Gravel at first.

\n\n

I found the instruction from zooniverse which is the template for volunteers to label the data quite simple to understand. Moreover, they give us a hint of how labels are generated :

\n\n

Fish & Gravel

\n\n

\"fish

\n\n

Sugar & Flower

\n\n

\"sugar

\n\n

Hope this helps!

", "messages": [ { "id": "627493", "postDate": "09/16/2019 03:59:51", "content": "

Hi friends!\nJust in cases that somebody like me still gets confused on the definitions of cloud types.\nE.g. I didn't understand different between Sugar vs. Gravel at first.

\n\n

I found the instruction from zooniverse which is the template for volunteers to label the data quite simple to understand. Moreover, they give us a hint of how labels are generated :

\n\n

Fish & Gravel

\n\n

\"fish

\n\n

Sugar & Flower

\n\n

\"sugar

\n\n

Hope this helps!

", "rawMarkdown": "Hi friends!\nJust in cases that somebody like me still gets confused on the definitions of cloud types.\nE.g. I didn't understand different between Sugar vs. Gravel at first.\n\nI found the instruction from zooniverse which is the template for volunteers to label the data quite simple to understand. Moreover, they give us a hint of how labels are generated :\n\n### Fish & Gravel\n![fish gravel](https://i.ibb.co/0Z868W2/fish-gravel.png)\n\n### Sugar & Flower\n![sugar flower](https://i.ibb.co/gVdnxZX/sugar-flower.png)\n\nHope this helps!", "votes": null }, { "id": "627516", "postDate": "09/16/2019 04:55:21", "content": "

Very Informative....!!\nThanks @ratthachat

", "rawMarkdown": "Very Informative....!!\nThanks @ratthachat", "votes": null }, { "id": "627555", "postDate": "09/16/2019 05:59:15", "content": "

Great to know! Thanks, Jung! đź‘Ť đź‘Ť

", "rawMarkdown": "Great to know! Thanks, Jung! đź‘Ť đź‘Ť", "votes": null }, { "id": "627698", "postDate": "09/16/2019 09:14:38", "content": "

Thanks for sharing this @ratthachat . It was indeed tough to classify Gravel and sugar in each of the images if present. I shall explore zooniverse on their labeling and try to use it to get best results for the model.

", "rawMarkdown": "Thanks for sharing this @ratthachat . It was indeed tough to classify Gravel and sugar in each of the images if present. I shall explore zooniverse on their labeling and try to use it to get best results for the model.", "votes": null }, { "id": "628807", "postDate": "09/18/2019 00:09:58", "content": "

Thanks @ratthachat , this explanation is very clear about the labels, to be honest, I wasn't completely sure before this.

", "rawMarkdown": "Thanks @ratthachat , this explanation is very clear about the labels, to be honest, I wasn't completely sure before this.", "votes": null }, { "id": "628809", "postDate": "09/18/2019 00:25:00", "content": "

You might also find the just within the last days published article on the precursor of this event helpful. It has been a much smaller hand-labeling effort (900 images), but led to the determination of the different patterns of shallow convection (sugar, flower, gravel, fish) we are looking for in this competition.

", "rawMarkdown": "You might also find the just within the last days [published article](https://doi.org/10.1002/qj.3662) on the precursor of this event helpful. It has been a much smaller hand-labeling effort (900 images), but led to the determination of the different patterns of shallow convection (sugar, flower, gravel, fish) we are looking for in this competition.", "votes": null }, { "id": "628813", "postDate": "09/18/2019 00:32:04", "content": "

Thanks for sharing the useful info!!

", "rawMarkdown": "Thanks for sharing the useful info!!", "votes": null }, { "id": "628856", "postDate": "09/18/2019 03:11:06", "content": "

@observingclouds \nHi Hauke, thanks for the link to the article! Nevertheless, I couldn’t open the article at all in both my ipad and mobile phone.

\n\n

The pdf link I found is https://rmets.onlinelibrary.wiley.com/doi/pdf/10.1002/qj.3662

\n\n

After several attempts of reloading, the best I can do is to see only the first page.\nCan anybody open this article? Or can you help Hauke?

", "rawMarkdown": "observingclouds \nHi Hauke, thanks for the link to the article! Nevertheless, I couldn’t open the article at all in both my ipad and mobile phone. \n\nThe pdf link I found is https://rmets.onlinelibrary.wiley.com/doi/pdf/10.1002/qj.3662\n\nAfter several attempts of reloading, the best I can do is to see only the first page.\nCan anybody open this article? Or can you help Hauke?", "votes": null }, { "id": "628867", "postDate": "09/18/2019 03:54:16", "content": "

Hi @ratthachat, the article is with about 60MB quite large. Nevertheless, you should be able to access the whole article as it is open-access.\nTry to use their epdf-reader and click then on the PDF download. For me it works with Safari and Firefox on Mac.

", "rawMarkdown": "Hi @ratthachat, the article is with about 60MB quite large. Nevertheless, you should be able to access the whole article as it is open-access.\nTry to use their [epdf-reader](https://doi.org/10.1002/qj.3662) and click then on the PDF download. For me it works with Safari and Firefox on Mac.", "votes": null }, { "id": "639496", "postDate": "10/03/2019 09:27:35", "content": "

Thanks for sharing. BTW, I tried hand labeling on zooniverse and I think it is a complex task at least if you are not expert in clouds. This is why I don't think very high accuracy (or Dice coef or whatever) can be reached in this competition. We just train on the set which different people have marked in different ways.

", "rawMarkdown": "Thanks for sharing. BTW, I tried hand labeling on zooniverse and I think it is a complex task at least if you are not expert in clouds. This is why I don't think very high accuracy (or Dice coef or whatever) can be reached in this competition. We just train on the set which different people have marked in different ways.", "votes": null }, { "id": "639642", "postDate": "10/03/2019 12:28:35", "content": "

@demonplus Anna, I think so. And I am still thinking what should be an optimal logic for this inherent noisy labeling process...

", "rawMarkdown": "demonplus Anna, I think so. And I am still thinking what should be an optimal logic for this inherent noisy labeling process...", "votes": null }, { "id": "659349", "postDate": "10/27/2019 13:50:44", "content": "

Thanks for the descriptions. When looking at the training data, can we see the names of the annotators, dates of annotation, and location?

", "rawMarkdown": "Thanks for the descriptions. When looking at the training data, can we see the names of the annotators, dates of annotation, and location?", "votes": null }, { "id": "659382", "postDate": "10/27/2019 14:55:35", "content": "

No annotator information is provided. You get the binary union of all annotators for the image. You can work out the locations but this information didn’t help me at least.

", "rawMarkdown": "No annotator information is provided. You get the binary union of all annotators for the image. You can work out the locations but this information didn’t help me at least.", "votes": null }, { "id": "659547", "postDate": "10/27/2019 21:51:51", "content": "

This is great. very clear for the pattens and labels. I was so confused with suger vs gravel also.\nVery helpful to a late joiner. Thanks a lot!

", "rawMarkdown": "This is great. very clear for the pattens and labels. I was so confused with suger vs gravel also.\nVery helpful to a late joiner. Thanks a lot!", "votes": null }, { "id": "665747", "postDate": "11/05/2019 10:58:41", "content": "

Very useful thanks

", "rawMarkdown": "Very useful thanks", "votes": null } ], "comments": [ { "id": 627516, "author_name": "veeralakrishna", "author_url": "", "post_date": "09/16/2019 04:55:21", "content": "

Very Informative....!!\nThanks @ratthachat

", "votes": null, "replies": [] }, { "id": 627555, "author_name": "samusram", "author_url": "", "post_date": "09/16/2019 05:59:15", "content": "

Great to know! Thanks, Jung! đź‘Ť đź‘Ť

", "votes": null, "replies": [] }, { "id": 627698, "author_name": "manojprabhaakr", "author_url": "", "post_date": "09/16/2019 09:14:38", "content": "

Thanks for sharing this @ratthachat . It was indeed tough to classify Gravel and sugar in each of the images if present. I shall explore zooniverse on their labeling and try to use it to get best results for the model.

", "votes": null, "replies": [] }, { "id": 628807, "author_name": "dimitreoliveira", "author_url": "", "post_date": "09/18/2019 00:09:58", "content": "

Thanks @ratthachat , this explanation is very clear about the labels, to be honest, I wasn't completely sure before this.

", "votes": null, "replies": [] }, { "id": 628809, "author_name": "observingclouds", "author_url": "", "post_date": "09/18/2019 00:25:00", "content": "

You might also find the just within the last days published article on the precursor of this event helpful. It has been a much smaller hand-labeling effort (900 images), but led to the determination of the different patterns of shallow convection (sugar, flower, gravel, fish) we are looking for in this competition.

", "votes": null, "replies": [ { "id": 628856, "author_name": "ratthachat", "author_url": "", "post_date": "09/18/2019 03:11:06", "content": "

@observingclouds \nHi Hauke, thanks for the link to the article! Nevertheless, I couldn’t open the article at all in both my ipad and mobile phone.

\n\n

The pdf link I found is https://rmets.onlinelibrary.wiley.com/doi/pdf/10.1002/qj.3662

\n\n

After several attempts of reloading, the best I can do is to see only the first page.\nCan anybody open this article? Or can you help Hauke?

", "votes": null, "replies": [] }, { "id": 628867, "author_name": "observingclouds", "author_url": "", "post_date": "09/18/2019 03:54:16", "content": "

Hi @ratthachat, the article is with about 60MB quite large. Nevertheless, you should be able to access the whole article as it is open-access.\nTry to use their epdf-reader and click then on the PDF download. For me it works with Safari and Firefox on Mac.

", "votes": null, "replies": [] } ] }, { "id": 628813, "author_name": "yuta0001", "author_url": "", "post_date": "09/18/2019 00:32:04", "content": "

Thanks for sharing the useful info!!

", "votes": null, "replies": [] }, { "id": 639496, "author_name": "demonplus", "author_url": "", "post_date": "10/03/2019 09:27:35", "content": "

Thanks for sharing. BTW, I tried hand labeling on zooniverse and I think it is a complex task at least if you are not expert in clouds. This is why I don't think very high accuracy (or Dice coef or whatever) can be reached in this competition. We just train on the set which different people have marked in different ways.

", "votes": null, "replies": [ { "id": 639642, "author_name": "ratthachat", "author_url": "", "post_date": "10/03/2019 12:28:35", "content": "

@demonplus Anna, I think so. And I am still thinking what should be an optimal logic for this inherent noisy labeling process...

", "votes": null, "replies": [] } ] }, { "id": 659349, "author_name": "cdeotte", "author_url": "", "post_date": "10/27/2019 13:50:44", "content": "

Thanks for the descriptions. When looking at the training data, can we see the names of the annotators, dates of annotation, and location?

", "votes": null, "replies": [ { "id": 659382, "author_name": "robga", "author_url": "", "post_date": "10/27/2019 14:55:35", "content": "

No annotator information is provided. You get the binary union of all annotators for the image. You can work out the locations but this information didn’t help me at least.

", "votes": null, "replies": [] } ] }, { "id": 659547, "author_name": "cooleel", "author_url": "", "post_date": "10/27/2019 21:51:51", "content": "

This is great. very clear for the pattens and labels. I was so confused with suger vs gravel also.\nVery helpful to a late joiner. Thanks a lot!

", "votes": null, "replies": [] }, { "id": 665747, "author_name": "erdeneochir", "author_url": "", "post_date": "11/05/2019 10:58:41", "content": "

Very useful thanks

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "627493": "Hi friends!\nJust in cases that somebody like me still gets confused on the definitions of cloud types.\nE.g. I didn't understand different between Sugar vs. Gravel at first.\n\nI found the instruction from zooniverse which is the template for volunteers to label the data quite simple to understand. Moreover, they give us a hint of how labels are generated :\n\n### Fish & Gravel\n![fish gravel](https://i.ibb.co/0Z868W2/fish-gravel.png)\n\n### Sugar & Flower\n![sugar flower](https://i.ibb.co/gVdnxZX/sugar-flower.png)\n\nHope this helps!", "627516": "Very Informative....!!\nThanks @ratthachat", "627555": "Great to know! Thanks, Jung! 👍 👍", "627698": "Thanks for sharing this @ratthachat . It was indeed tough to classify Gravel and sugar in each of the images if present. I shall explore zooniverse on their labeling and try to use it to get best results for the model.", "628807": "Thanks @ratthachat , this explanation is very clear about the labels, to be honest, I wasn't completely sure before this.", "628809": "You might also find the just within the last days [published article](https://doi.org/10.1002/qj.3662) on the precursor of this event helpful. It has been a much smaller hand-labeling effort (900 images), but led to the determination of the different patterns of shallow convection (sugar, flower, gravel, fish) we are looking for in this competition.", "628813": "Thanks for sharing the useful info!!", "628856": "observingclouds \nHi Hauke, thanks for the link to the article! Nevertheless, I couldn’t open the article at all in both my ipad and mobile phone. \n\nThe pdf link I found is https://rmets.onlinelibrary.wiley.com/doi/pdf/10.1002/qj.3662\n\nAfter several attempts of reloading, the best I can do is to see only the first page.\nCan anybody open this article? Or can you help Hauke?", "628867": "Hi @ratthachat, the article is with about 60MB quite large. Nevertheless, you should be able to access the whole article as it is open-access.\nTry to use their [epdf-reader](https://doi.org/10.1002/qj.3662) and click then on the PDF download. For me it works with Safari and Firefox on Mac.", "639496": "Thanks for sharing. BTW, I tried hand labeling on zooniverse and I think it is a complex task at least if you are not expert in clouds. This is why I don't think very high accuracy (or Dice coef or whatever) can be reached in this competition. We just train on the set which different people have marked in different ways.", "639642": "demonplus Anna, I think so. And I am still thinking what should be an optimal logic for this inherent noisy labeling process...", "659349": "Thanks for the descriptions. When looking at the training data, can we see the names of the annotators, dates of annotation, and location?", "659382": "No annotator information is provided. You get the binary union of all annotators for the image. You can work out the locations but this information didn’t help me at least.", "659547": "This is great. very clear for the pattens and labels. I was so confused with suger vs gravel also.\nVery helpful to a late joiner. Thanks a lot!", "665747": "Very useful thanks" }, "source": "meta" }