![\"https://storage.googleapis.com/kaggle-competitions/kaggle/31703/logos/header.png\"](\"https://storage.googleapis.com/kaggle-competitions/kaggle/31703/logos/header.png\")
this is the one:
\n
interesting, if you search image via google lens, it correctly identify this image as COTS
", "messages": [ { "id": "1644885", "postDate": "01/10/2022 15:24:56", "content": "this is the one:
\n
interesting, if you search image via google lens, it correctly identify this image as COTS
", "rawMarkdown": "this is the one:\n\n\n\n\ninteresting, if you search image via google lens, it correctly identify this image as COTS", "votes": null }, { "id": "1644890", "postDate": "01/10/2022 15:27:48", "content": "Thank you! Hope … our score goes up by ….. 0.0001 :) and everybody jump on leaderboard …. let's see :)
", "rawMarkdown": "Thank you! Hope ... our score goes up by ..... 0.0001 :) and everybody jump on leaderboard .... let's see :)", "votes": null }, { "id": "1645164", "postDate": "01/10/2022 18:17:08", "content": "The secret sauce to the top LB submissions 😂😂😂
", "rawMarkdown": "The secret sauce to the top LB submissions 😂😂😂", "votes": null }, { "id": "1645196", "postDate": "01/10/2022 18:43:18", "content": "how to blend this into current video data:
\nhttps://openaccess.thecvf.com/content_WACV_2020/papers/Zhang_Deep_Image_Blending_WACV_2020_paper.pdf
didn't know that you can buy cots on ebay : https://www.ebay.com/itm/224430306583?
", "rawMarkdown": "didn't know that you can buy cots on ebay : https://www.ebay.com/itm/224430306583?", "votes": null }, { "id": "1645743", "postDate": "01/11/2022 07:19:39", "content": "It is time to travel south and take some underwater video😂
", "rawMarkdown": "It is time to travel south and take some underwater video😂", "votes": null }, { "id": "1645805", "postDate": "01/11/2022 08:27:58", "content": "We can use google image to download COTS image and use it as this one
", "rawMarkdown": "We can use google image to download COTS image and use it as this one", "votes": null }, { "id": "1646673", "postDate": "01/11/2022 23:05:22", "content": "not sure if you need to check their license?
", "rawMarkdown": "not sure if you need to check their license?", "votes": null }, { "id": "1646764", "postDate": "01/12/2022 02:55:05", "content": "Well, I think images from google are \"publicly available\" :D , maybe It better to be confirmed by the host
", "rawMarkdown": "Well, I think images from google are \"publicly available\" :D , maybe It better to be confirmed by the host", "votes": null }, { "id": "1646928", "postDate": "01/12/2022 07:19:09", "content": "One image and 3 starfishes so beauty. Current dataset is very limited unfortunately. Data augumentation is the key.
", "rawMarkdown": "One image and 3 starfishes so beauty. Current dataset is very limited unfortunately. Data augumentation is the key.", "votes": null }, { "id": "1646975", "postDate": "01/12/2022 08:08:37", "content": "You can only use any extra data with commercial license for this competition.
", "rawMarkdown": "You can only use any extra data with commercial license for this competition.", "votes": null }, { "id": "1647284", "postDate": "01/12/2022 13:09:44", "content": "i thought we can use \"Creative Commons licenses\"?
\nif any kaggler can confirm this, I can collect and share such a dataset.
(on a side note, I find it easier to experiment with augmentation by using a say 64x64 image classifier on cropped COTS images from the kaggle dataset. this can be used to pretrain the backbone of your object detection net and also as a post-process verifier)
\nI am experimenting with GAN inpainting. the idea is to draw a black square mask on kaggle dataset and then the GAN to inpaint a COTS in it
\n![\"https://i.ibb.co/nDmydc6/Selection-999-710.png\"](\"https://i.ibb.co/nDmydc6/Selection-999-710.png\")
i wonder if i mixup sea urchin + sunflower starfish would work (since CC license of COTSS images are limited)
", "rawMarkdown": "i thought we can use \"Creative Commons licenses\"?\nif any kaggler can confirm this, I can collect and share such a dataset.\n\n(on a side note, I find it easier to experiment with augmentation by using a say 64x64 image classifier on cropped COTS images from the kaggle dataset. this can be used to pretrain the backbone of your object detection net and also as a post-process verifier)\n\nI am experimenting with GAN inpainting. the idea is to draw a black square mask on kaggle dataset and then the GAN to inpaint a COTS in it\n\n\n\n\n i wonder if i mixup sea urchin + sunflower starfish would work (since CC license of COTSS images are limited)", "votes": null }, { "id": "1647316", "postDate": "01/12/2022 13:42:16", "content": "Sure, Creative Commons is OK, it allows commercial usage.
", "rawMarkdown": "Sure, Creative Commons is OK, it allows commercial usage.", "votes": null }, { "id": "1647390", "postDate": "01/12/2022 14:59:41", "content": "there are 4 cots
", "rawMarkdown": "there are 4 cots", "votes": null }, { "id": "1647563", "postDate": "01/12/2022 17:34:10", "content": "@ptran1203 Google results will return images with a variety of licenses, but as @hengck23 noted you can restrict the results licenses to creative commons by specifying that it's an image search, clicking tools, and then selecting \"usage rights\". I don't know exactly what goes on under the hood there so I can't promise it will literally never return images that violate the rules but I would be surprised if that's at all common.
", "rawMarkdown": "ptran1203 Google results will return images with a variety of licenses, but as @hengck23 noted you can restrict the results licenses to creative commons by specifying that it's an image search, clicking tools, and then selecting \"usage rights\". I don't know exactly what goes on under the hood there so I can't promise it will literally never return images that violate the rules but I would be surprised if that's at all common.", "votes": null }, { "id": "1647659", "postDate": "01/12/2022 18:59:51", "content": "One false negative, it is ok
", "rawMarkdown": "One false negative, it is ok", "votes": null }, { "id": "1647915", "postDate": "01/13/2022 02:06:46", "content": "Unexpected tips.！！！ I hope the image can bring a little improvement . Hahahahahahahahahaha😂😂😂
", "rawMarkdown": "Unexpected tips.！！！ I hope the image can bring a little improvement . Hahahahahahahahahaha😂😂😂", "votes": null }, { "id": "1648282", "postDate": "01/13/2022 10:01:40", "content": "Just wondering if I bought the COTS and took a lot of pictures, is it illegal to train using those pictures?
", "rawMarkdown": "Just wondering if I bought the COTS and took a lot of pictures, is it illegal to train using those pictures?", "votes": null }, { "id": "1648288", "postDate": "01/13/2022 10:05:01", "content": "@onodera let's go to Australia and take picture :)
", "rawMarkdown": "onodera let's go to Australia and take picture :)", "votes": null }, { "id": "1648298", "postDate": "01/13/2022 10:15:09", "content": "Another way to use the google image without license issue is \"Human GAN\":
\n@remekkinas think COTSs are still living in Okinawa, south island of Japan. But I don't want to dive in this season 🤒
", "rawMarkdown": "remekkinas think COTSs are still living in Okinawa, south island of Japan. But I don't want to dive in this season 🤒", "votes": null }, { "id": "1654361", "postDate": "01/18/2022 12:18:45", "content": "@onoderamasato
\ni think you don't have to go to okinawa …. you can make your own stafish
http://www.xfallenmoon.com/blog/diy-how-to-make-faux-starfish
\n![\"https://i.ibb.co/HVVWfm1/How-to-make-faux-starfish-with-clay-Super-easy-step-by-step-on-how-to-create-faux-starfish-from-scra.jpg\"](\"https://i.ibb.co/HVVWfm1/How-to-make-faux-starfish-with-clay-Super-easy-step-by-step-on-how-to-create-faux-starfish-from-scra.jpg\")
![\"https://1.bp.blogspot.com/-ZS0uEEliiEM/U_uw-gd5xCI/AAAAAAABALM/C9jXvDslwhM/s1600/clay-starfish-DIY.jpg\"](\"https://1.bp.blogspot.com/-ZS0uEEliiEM/U_uw-gd5xCI/AAAAAAABALM/C9jXvDslwhM/s1600/clay-starfish-DIY.jpg\")
interesting. Thank you for sharing a picture!
", "rawMarkdown": "interesting. Thank you for sharing a picture!", "votes": null }, { "id": "1657414", "postDate": "01/20/2022 07:01:03", "content": "You can download the official video (https://www.youtube.com/watch?v=UT2noVDFoaA). There are many starfishes and maybe some test images are from it.
", "rawMarkdown": "You can download the official video (https://www.youtube.com/watch?v=UT2noVDFoaA). There are many starfishes and maybe some test images are from it.", "votes": null }, { "id": "1657419", "postDate": "01/20/2022 07:03:28", "content": "\"maybe some test images are from it.\"
\ni have check it. the host are careful not to leaked any test images. the anootated video in the youtube are from video0 and video2
", "rawMarkdown": "\"maybe some test images are from it.\"\n\ni have check it. the host are careful not to leaked any test images. the anootated video in the youtube are from video0 and video2", "votes": null }, { "id": "1657430", "postDate": "01/20/2022 07:09:06", "content": "Thank you for your reply!
", "rawMarkdown": "Thank you for your reply!", "votes": null }, { "id": "1660023", "postDate": "01/22/2022 10:23:43", "content": "Wow, amazing discovery. Thanks, I think now we all can get a better model, a huge step toward 0.8 LB.
", "rawMarkdown": "Wow, amazing discovery. Thanks, I think now we all can get a better model, a huge step toward 0.8 LB.", "votes": null } ], "comments": [ { "id": 1644890, "author_name": "remekkinas", "author_url": "", "post_date": "01/10/2022 15:27:48", "content": "Thank you! Hope … our score goes up by ….. 0.0001 :) and everybody jump on leaderboard …. let's see :)
", "votes": null, "replies": [] }, { "id": 1645164, "author_name": "maxvandijck", "author_url": "", "post_date": "01/10/2022 18:17:08", "content": "The secret sauce to the top LB submissions 😂😂😂
", "votes": null, "replies": [] }, { "id": 1645196, "author_name": "hengck23", "author_url": "", "post_date": "01/10/2022 18:43:18", "content": "how to blend this into current video data:
\nhttps://openaccess.thecvf.com/content_WACV_2020/papers/Zhang_Deep_Image_Blending_WACV_2020_paper.pdf
didn't know that you can buy cots on ebay : https://www.ebay.com/itm/224430306583?
", "votes": null, "replies": [ { "id": 1645743, "author_name": "steamedsheep", "author_url": "", "post_date": "01/11/2022 07:19:39", "content": "It is time to travel south and take some underwater video😂
", "votes": null, "replies": [] } ] }, { "id": 1645805, "author_name": "ptran1203", "author_url": "", "post_date": "01/11/2022 08:27:58", "content": "We can use google image to download COTS image and use it as this one
", "votes": null, "replies": [ { "id": 1646673, "author_name": "hengck23", "author_url": "", "post_date": "01/11/2022 23:05:22", "content": "not sure if you need to check their license?
", "votes": null, "replies": [] }, { "id": 1646764, "author_name": "ptran1203", "author_url": "", "post_date": "01/12/2022 02:55:05", "content": "Well, I think images from google are \"publicly available\" :D , maybe It better to be confirmed by the host
", "votes": null, "replies": [] }, { "id": 1646975, "author_name": "philippsinger", "author_url": "", "post_date": "01/12/2022 08:08:37", "content": "You can only use any extra data with commercial license for this competition.
", "votes": null, "replies": [] }, { "id": 1647284, "author_name": "hengck23", "author_url": "", "post_date": "01/12/2022 13:09:44", "content": "i thought we can use \"Creative Commons licenses\"?
\nif any kaggler can confirm this, I can collect and share such a dataset.
(on a side note, I find it easier to experiment with augmentation by using a say 64x64 image classifier on cropped COTS images from the kaggle dataset. this can be used to pretrain the backbone of your object detection net and also as a post-process verifier)
\nI am experimenting with GAN inpainting. the idea is to draw a black square mask on kaggle dataset and then the GAN to inpaint a COTS in it
\n
i wonder if i mixup sea urchin + sunflower starfish would work (since CC license of COTSS images are limited)
", "votes": null, "replies": [] }, { "id": 1647316, "author_name": "philippsinger", "author_url": "", "post_date": "01/12/2022 13:42:16", "content": "Sure, Creative Commons is OK, it allows commercial usage.
", "votes": null, "replies": [] }, { "id": 1647563, "author_name": "sohier", "author_url": "", "post_date": "01/12/2022 17:34:10", "content": "@ptran1203 Google results will return images with a variety of licenses, but as @hengck23 noted you can restrict the results licenses to creative commons by specifying that it's an image search, clicking tools, and then selecting \"usage rights\". I don't know exactly what goes on under the hood there so I can't promise it will literally never return images that violate the rules but I would be surprised if that's at all common.
", "votes": null, "replies": [] }, { "id": 1648282, "author_name": "onodera", "author_url": "", "post_date": "01/13/2022 10:01:40", "content": "Just wondering if I bought the COTS and took a lot of pictures, is it illegal to train using those pictures?
", "votes": null, "replies": [] }, { "id": 1648288, "author_name": "remekkinas", "author_url": "", "post_date": "01/13/2022 10:05:01", "content": "@onodera let's go to Australia and take picture :)
", "votes": null, "replies": [] }, { "id": 1648329, "author_name": "onodera", "author_url": "", "post_date": "01/13/2022 11:02:11", "content": "@remekkinas think COTSs are still living in Okinawa, south island of Japan. But I don't want to dive in this season 🤒
", "votes": null, "replies": [] } ] }, { "id": 1646928, "author_name": "marcinstasko", "author_url": "", "post_date": "01/12/2022 07:19:09", "content": "One image and 3 starfishes so beauty. Current dataset is very limited unfortunately. Data augumentation is the key.
", "votes": null, "replies": [ { "id": 1647390, "author_name": "hengck23", "author_url": "", "post_date": "01/12/2022 14:59:41", "content": "there are 4 cots
", "votes": null, "replies": [] }, { "id": 1647659, "author_name": "speedwagon", "author_url": "", "post_date": "01/12/2022 18:59:51", "content": "One false negative, it is ok
", "votes": null, "replies": [] } ] }, { "id": 1647915, "author_name": "aihouhaoxiong", "author_url": "", "post_date": "01/13/2022 02:06:46", "content": "Unexpected tips.！！！ I hope the image can bring a little improvement . Hahahahahahahahahaha😂😂😂
", "votes": null, "replies": [] }, { "id": 1648298, "author_name": "hengck23", "author_url": "", "post_date": "01/13/2022 10:15:09", "content": "Another way to use the google image without license issue is \"Human GAN\":
\n@onoderamasato
\ni think you don't have to go to okinawa …. you can make your own stafish
http://www.xfallenmoon.com/blog/diy-how-to-make-faux-starfish
\n
interesting. Thank you for sharing a picture!
", "votes": null, "replies": [] }, { "id": 1657414, "author_name": "xinjiali", "author_url": "", "post_date": "01/20/2022 07:01:03", "content": "You can download the official video (https://www.youtube.com/watch?v=UT2noVDFoaA). There are many starfishes and maybe some test images are from it.
", "votes": null, "replies": [ { "id": 1657419, "author_name": "hengck23", "author_url": "", "post_date": "01/20/2022 07:03:28", "content": "\"maybe some test images are from it.\"
\ni have check it. the host are careful not to leaked any test images. the anootated video in the youtube are from video0 and video2
", "votes": null, "replies": [] }, { "id": 1657430, "author_name": "xinjiali", "author_url": "", "post_date": "01/20/2022 07:09:06", "content": "Thank you for your reply!
", "votes": null, "replies": [] } ] }, { "id": 1660023, "author_name": "calvchen", "author_url": "", "post_date": "01/22/2022 10:23:43", "content": "Wow, amazing discovery. Thanks, I think now we all can get a better model, a huge step toward 0.8 LB.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1644885": "this is the one:\n\n\n\n\ninteresting, if you search image via google lens, it correctly identify this image as COTS", "1644890": "Thank you! Hope ... our score goes up by ..... 0.0001 :) and everybody jump on leaderboard .... let's see :)", "1645164": "The secret sauce to the top LB submissions 😂😂😂", "1645196": "how to blend this into current video data:\nhttps://openaccess.thecvf.com/content_WACV_2020/papers/Zhang_Deep_Image_Blending_WACV_2020_paper.pdf", "1645316": "didn't know that you can buy cots on ebay : https://www.ebay.com/itm/224430306583?", "1645743": "It is time to travel south and take some underwater video😂", "1645805": "We can use google image to download COTS image and use it as this one", "1646673": "not sure if you need to check their license?", "1646764": "Well, I think images from google are \"publicly available\" :D , maybe It better to be confirmed by the host", "1646928": "One image and 3 starfishes so beauty. Current dataset is very limited unfortunately. Data augumentation is the key.", "1646975": "You can only use any extra data with commercial license for this competition.", "1647284": "i thought we can use \"Creative Commons licenses\"?\nif any kaggler can confirm this, I can collect and share such a dataset.\n\n(on a side note, I find it easier to experiment with augmentation by using a say 64x64 image classifier on cropped COTS images from the kaggle dataset. this can be used to pretrain the backbone of your object detection net and also as a post-process verifier)\n\nI am experimenting with GAN inpainting. the idea is to draw a black square mask on kaggle dataset and then the GAN to inpaint a COTS in it\n\n\n\n\n i wonder if i mixup sea urchin + sunflower starfish would work (since CC license of COTSS images are limited)", "1647316": "Sure, Creative Commons is OK, it allows commercial usage.", "1647390": "there are 4 cots", "1647563": "ptran1203 Google results will return images with a variety of licenses, but as @hengck23 noted you can restrict the results licenses to creative commons by specifying that it's an image search, clicking tools, and then selecting \"usage rights\". I don't know exactly what goes on under the hood there so I can't promise it will literally never return images that violate the rules but I would be surprised if that's at all common.", "1647659": "One false negative, it is ok", "1647915": "Unexpected tips.！！！ I hope the image can bring a little improvement . Hahahahahahahahahaha😂😂😂", "1648282": "Just wondering if I bought the COTS and took a lot of pictures, is it illegal to train using those pictures?", "1648288": "onodera let's go to Australia and take picture :)", "1648298": "Another way to use the google image without license issue is \"Human GAN\":\n- open up your photoshop and draw the cots (assuming that you are a good artist)", "1648329": "remekkinas think COTSs are still living in Okinawa, south island of Japan. But I don't want to dive in this season 🤒", "1654361": "onoderamasato \ni think you don't have to go to okinawa .... you can make your own stafish\n\nhttp://www.xfallenmoon.com/blog/diy-how-to-make-faux-starfish\n\n\n\n", "1657148": "interesting. Thank you for sharing a picture!", "1657414": "You can download the official video (https://www.youtube.com/watch?v=UT2noVDFoaA). There are many starfishes and maybe some test images are from it.", "1657419": "\"maybe some test images are from it.\"\n\ni have check it. the host are careful not to leaked any test images. the anootated video in the youtube are from video0 and video2", "1657430": "Thank you for your reply!", "1660023": "Wow, amazing discovery. Thanks, I think now we all can get a better model, a huge step toward 0.8 LB." }, "source": "meta" }