![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F479538%2Fbbebf07caaefcc7407ad0ad52f7ff375%2Flist.PNG?generation=1562559016340484&alt=media\")
Welcome to the APTOS 2019 Blindness Detection challenge! It's been several years since Kaggle ran a similar competition, so we're excited to see what the progress in ML techniques can contribute to this problem. This time we have teamed up with the Aravind Eye Care Center in India which is committed to using AI to help individuals in underserved areas.
\n\nWe've retained as much data as possible for the test set to ensure the leaderboard scores are meaningful, so the train set is relatively small. You may wish to take advantage of the external data rules to incorporate other diabetic retinopathy datasets into your training set. Just don't forget to declare your sources in the external data thread
", "messages": [ { "id": 563162, "postDate": "2019-06-28T01:17:03.910Z", "content": "Welcome to the APTOS 2019 Blindness Detection challenge! It's been several years since Kaggle ran a similar competition, so we're excited to see what the progress in ML techniques can contribute to this problem. This time we have teamed up with the Aravind Eye Care Center in India which is committed to using AI to help individuals in underserved areas.
\n\nWe've retained as much data as possible for the test set to ensure the leaderboard scores are meaningful, so the train set is relatively small. You may wish to take advantage of the external data rules to incorporate other diabetic retinopathy datasets into your training set. Just don't forget to declare your sources in the external data thread
", "rawMarkdown": "Welcome to the APTOS 2019 Blindness Detection challenge! It's been several years since Kaggle ran a similar competition, so we're excited to see what the progress in ML techniques can contribute to this problem. This time we have teamed up with the Aravind Eye Care Center in India which is committed to using AI to help individuals in underserved areas.\n\nWe've retained as much data as possible for the test set to ensure the leaderboard scores are meaningful, so the train set is relatively small. You may wish to take advantage of the external data rules to incorporate other diabetic retinopathy datasets into your training set. Just don't forget to declare your sources in the external data thread\n", "votes": 31 }, { "id": 570261, "postDate": "2019-07-08T04:12:06.560Z", "content": "Hi @sohier , thanks for organizing this competition.
\nAs discussed in another thread, there are duplicated images in train and test.
I wrote up a simple kernel folked from @manojprabhaakr 's original one.
\nhttps://www.kaggle.com/maxwell110/duplicated-list-csv-file/output
After I checked through the duplicated list, I confirmed some duplicated images have different labels.
\n\nexamples\n
How should we handle this ?
\n\nThanks in advance.
", "rawMarkdown": "Hi @sohier , thanks for organizing this competition. \nAs discussed in [another thread](https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/98856#latest-570171), there are duplicated images in train and test. \n\nI wrote up a simple kernel folked from @manojprabhaakr 's original one. \nhttps://www.kaggle.com/maxwell110/duplicated-list-csv-file/output \n \nAfter I checked through the duplicated list, I confirmed some duplicated images have different labels. \n\n**examples**\n\n\n\nHow should we handle this ?\n\nThanks in advance.", "votes": 9, "replies": [ { "id": 571927, "postDate": "2019-07-10T08:42:16.687Z", "content": "I remove the duplicated image with different label
", "rawMarkdown": "I remove the duplicated image with different label" }, { "id": 573105, "postDate": "2019-07-11T19:33:34.647Z", "content": "@dexter6855 \nYes. I also think removing the duplicated images is one reasonable way, although not sure that makes our models better or worse :(\nNow I'm afraid that there might be duplicated images in private test with different labels...
", "rawMarkdown": "@dexter6855 \nYes. I also think removing the duplicated images is one reasonable way, although not sure that makes our models better or worse :(\nNow I'm afraid that there might be duplicated images in **private test** with different labels... ", "votes": 2 }, { "id": 573385, "postDate": "2019-07-12T07:39:10.697Z", "content": "I agree with you, it's looks like diagnosis for one case from various doctors, and our main task is generalised their decision. So for my mind more robust decision must use various opinions.
", "rawMarkdown": "I agree with you, it's looks like diagnosis for one case from various doctors, and our main task is generalised their decision. So for my mind more robust decision must use various opinions.", "votes": 2 }, { "id": 573397, "postDate": "2019-07-12T08:09:51.910Z", "content": "@miklgr500 \nYes...\nConsidering about that, I'm suspicious of excluding the duplicated images.\nEverything will depend on the property of private test data what we can not know.
", "rawMarkdown": "@miklgr500 \nYes...\nConsidering about that, I'm suspicious of excluding the duplicated images.\nEverything will depend on the property of private test data what we can not know." } ] }, { "id": 570046, "postDate": "2019-07-07T17:54:59.397Z", "content": "As I understood, kernels are run during submission with unseen test dataset. Yes? And public leaderboard is calculated using 15% of unseen test dataset? If it is a true, then why do we need current test dataset?
", "rawMarkdown": "As I understood, kernels are run during submission with unseen test dataset. Yes? And public leaderboard is calculated using 15% of unseen test dataset? If it is a true, then why do we need current test dataset?", "votes": 2 }, { "id": 565667, "postDate": "2019-07-01T08:55:20.937Z", "content": "Maybe close the previous competition's submission entrance is better.
", "rawMarkdown": "Maybe close the previous competition's submission entrance is better.", "votes": 2 }, { "id": 565650, "postDate": "2019-07-01T08:27:10.040Z", "content": "Thank you for hosting this competition.
\nBut I and some kagglers suffered submission error.
\nHere is my post.
\nhttps://www.kaggle.com/c/aptos2019-blindness-detection/discussion/98058#latest-565488
\nI would like to show the error log to us when submission fail.
\nThis will save our precious time:)
Please see the kernels requirements tab:
\n\nIn synchronous KO competitions, note that a submission that results in an error--either within the kernel or within the process of scoring--will count against your daily submission limit and will not return the specific error message. This is necessary to prevent probing the private test set.
Hi Sohier,
\n\nCan we train the model locally and use the kernel for infer only, just like the imet competition?
", "rawMarkdown": "Hi Sohier,\n\nCan we train the model locally and use the kernel for infer only, just like the imet competition?", "votes": 2, "replies": [ { "id": 563347, "postDate": "2019-06-28T07:18:24.957Z", "content": "I think you can.\nCheck the kernel requirements.\nhttps://www.kaggle.com/c/aptos2019-blindness-detection/overview/kernels-requirements
\n\n\n", "rawMarkdown": "I think you can.\nCheck the kernel requirements.\nhttps://www.kaggle.com/c/aptos2019-blindness-detection/overview/kernels-requirements\n\n> You can still train a model offline, upload it as a dataset, and use the kernel exclusively to perform inference.", "votes": 1 }, { "id": 563582, "postDate": "2019-06-28T13:05:01.310Z", "content": "You can still train a model offline, upload it as a dataset, and use the kernel exclusively to perform inference.
\n
Thanks Maxwell
", "rawMarkdown": "Thanks Maxwell", "votes": 1 }, { "id": 563709, "postDate": "2019-06-28T15:39:34.340Z", "content": "Yes, Maxwell is correct.
", "rawMarkdown": "Yes, Maxwell is correct." }, { "id": 563924, "postDate": "2019-06-28T19:30:30.593Z", "content": "Hi @sohier, just to clarify: if I train my model offline and upload it, I can keep it as a private dataset, right?
", "rawMarkdown": "Hi @sohier, just to clarify: if I train my model offline and upload it, I can keep it as a private dataset, right?", "votes": 1 }, { "id": 564255, "postDate": "2019-06-29T07:51:59.660Z", "content": "Yes you can do that.
", "rawMarkdown": "Yes you can do that.", "votes": 1 }, { "id": 565975, "postDate": "2019-07-01T16:30:59.717Z", "content": "@kozodoi yes, you can keep your trained model as a private dataset.
", "rawMarkdown": "@kozodoi yes, you can keep your trained model as a private dataset. ", "votes": 1 } ] }, { "id": 1345596, "postDate": "2021-06-11T17:27:51.863Z", "content": "Hi Sohier
\nI'm currently working as a PhD student working with retinal images and I would like to know if you could publish the solution file (labels of the test images) for this competition, since it already finished.
We do not have plans to release the solution file so that we can preserve the test set for future opportunities.
", "rawMarkdown": "We do not have plans to release the solution file so that we can preserve the test set for future opportunities. " } ] }, { "id": 1101541, "postDate": "2020-12-04T02:11:35.157Z", "content": "Hi @sohier,
\nDo you know if the '0 - No DR' class contains images of healthy/normal retina, or could these images show other ophthalmic pathologies?
\nThank you very much.
", "rawMarkdown": "Hi @sohier,\n\nDo you know if the '0 - No DR' class contains images of healthy/normal retina, or could these images show other ophthalmic pathologies?\n\nThank you very much." }, { "id": 622110, "postDate": "2019-09-09T09:31:26.957Z", "content": "@sohier are you going to release the public test set solution file?
", "rawMarkdown": "@sohier are you going to release the public test set solution file? " }, { "id": 607638, "postDate": "2019-08-25T16:32:18.437Z", "content": "Why is it so slow when I 'Select Files to Upload'? 15MB of data took about 1h.
", "rawMarkdown": "Why is it so slow when I 'Select Files to Upload'? 15MB of data took about 1h." }, { "id": 580320, "postDate": "2019-07-20T02:11:43.343Z", "content": "Hi @sohier , I have a question please.
\n\nIn this synchonous competition, since the submission is already calculated on private dataset, is it safe to delete/update some datasets which produce that score?
\n\nExample. I trained my model with dataset1 and got 0.777. After that, if I accidentally update (or delete) dataset1 (and it will affect the training process), will this affect the final score?
", "rawMarkdown": "Hi @sohier , I have a question please.\n\nIn this synchonous competition, since the submission is already calculated on private dataset, is it safe to delete/update some datasets which produce that score?\n\nExample. I trained my model with dataset1 and got 0.777. After that, if I accidentally update (or delete) dataset1 (and it will affect the training process), will this affect the final score?" }, { "id": 569912, "postDate": "2019-07-07T14:48:18.623Z", "content": "Hi Sohier!
\n\n\n\n\nYour kernel will re-run automatically against an unseen test set
\n
How size of this \"unseen\" test set is differ relative to current test set? I think we need this information to fit to execution time limit.
", "rawMarkdown": "Hi Sohier!\n\n> Your kernel will re-run automatically against an unseen test set\n\nHow size of this \"unseen\" test set is differ relative to current test set? I think we need this information to fit to execution time limit.", "replies": [ { "id": 570086, "postDate": "2019-07-07T18:54:28.170Z", "content": "Found the answer on the Data page. It will be about 13k images.
", "rawMarkdown": "Found the answer on the Data page. It will be about 13k images." } ] }, { "id": 567822, "postDate": "2019-07-04T03:40:37.333Z", "content": "Hi Sohier,
\n\nCan we use custom packages that are not pre-installed in the kernel?
", "rawMarkdown": "Hi Sohier, \n\nCan we use custom packages that are not pre-installed in the kernel? ", "replies": [ { "id": 568919, "postDate": "2019-07-05T16:57:32.483Z", "content": "Yes, as long as those packages comply with the rest of the rules like the requirement for winning solutions to be open sourced.
", "rawMarkdown": "Yes, as long as those packages comply with the rest of the rules like the requirement for winning solutions to be open sourced." }, { "id": 570159, "postDate": "2019-07-07T23:09:14.477Z", "content": "Thanks for the clarification.
", "rawMarkdown": "Thanks for the clarification. " } ] }, { "id": 566609, "postDate": "2019-07-02T11:06:57.670Z", "content": "Sounds like a good use case. Thanks for hosting this competition. @sohier
", "rawMarkdown": "Sounds like a good use case. Thanks for hosting this competition. @sohier " }, { "id": 565183, "postDate": "2019-06-30T14:49:22.030Z", "content": "Sounds like a great use case for ML! I will try to get a team together for this!
", "rawMarkdown": "Sounds like a great use case for ML! I will try to get a team together for this!" }, { "id": 564506, "postDate": "2019-06-29T14:16:25.663Z", "content": "Sounds exciting !!, will participate in this.
", "rawMarkdown": "Sounds exciting !!, will participate in this." }, { "id": 564069, "postDate": "2019-06-29T00:07:31.353Z", "content": "Hi @sohier, it's currently 00:06 29/06 UTC as I'm typing this, however my submission count for today hasn't been reset and I'll have to wait for 24 more hours. Is this working as intended?
", "rawMarkdown": "Hi @sohier, it's currently 00:06 29/06 UTC as I'm typing this, however my submission count for today hasn't been reset and I'll have to wait for 24 more hours. Is this working as intended?" }, { "id": 566063, "postDate": "2019-07-01T18:56:44.397Z", "rawMarkdown": "", "isDeleted": true }, { "id": 607649, "postDate": "2019-08-25T16:53:55.083Z", "content": "Thanks
", "rawMarkdown": "Thanks" } ], "comments": [ { "id": 570261, "author_name": "Maxwell", "author_url": "", "post_date": "2019-07-08T04:12:06.560000", "content": "Hi @sohier , thanks for organizing this competition.
\nAs discussed in another thread, there are duplicated images in train and test.
I wrote up a simple kernel folked from @manojprabhaakr 's original one.
\nhttps://www.kaggle.com/maxwell110/duplicated-list-csv-file/output
After I checked through the duplicated list, I confirmed some duplicated images have different labels.
\n\nexamples\n
How should we handle this ?
\n\nThanks in advance.
", "votes": 9, "replies": [ { "id": 571927, "author_name": "dexter6855", "author_url": "", "post_date": "2019-07-10T08:42:16.687000", "content": "I remove the duplicated image with different label
", "votes": 0, "replies": [] }, { "id": 573105, "author_name": "Maxwell", "author_url": "", "post_date": "2019-07-11T19:33:34.647000", "content": "@dexter6855 \nYes. I also think removing the duplicated images is one reasonable way, although not sure that makes our models better or worse :(\nNow I'm afraid that there might be duplicated images in private test with different labels...
", "votes": 2, "replies": [] }, { "id": 573385, "author_name": "Welf Crozzo", "author_url": "", "post_date": "2019-07-12T07:39:10.697000", "content": "I agree with you, it's looks like diagnosis for one case from various doctors, and our main task is generalised their decision. So for my mind more robust decision must use various opinions.
", "votes": 2, "replies": [] }, { "id": 573397, "author_name": "Maxwell", "author_url": "", "post_date": "2019-07-12T08:09:51.910000", "content": "@miklgr500 \nYes...\nConsidering about that, I'm suspicious of excluding the duplicated images.\nEverything will depend on the property of private test data what we can not know.
", "votes": 0, "replies": [] } ] }, { "id": 570046, "author_name": "Siarhei Fedartsou", "author_url": "", "post_date": "2019-07-07T17:54:59.397000", "content": "As I understood, kernels are run during submission with unseen test dataset. Yes? And public leaderboard is calculated using 15% of unseen test dataset? If it is a true, then why do we need current test dataset?
", "votes": 2, "replies": [] }, { "id": 565667, "author_name": "Yiheng Wang", "author_url": "", "post_date": "2019-07-01T08:55:20.937000", "content": "Maybe close the previous competition's submission entrance is better.
", "votes": 2, "replies": [] }, { "id": 565650, "author_name": "takuoko", "author_url": "", "post_date": "2019-07-01T08:27:10.040000", "content": "Thank you for hosting this competition.
\nBut I and some kagglers suffered submission error.
\nHere is my post.
\nhttps://www.kaggle.com/c/aptos2019-blindness-detection/discussion/98058#latest-565488
\nI would like to show the error log to us when submission fail.
\nThis will save our precious time:)
Please see the kernels requirements tab:
\n\nIn synchronous KO competitions, note that a submission that results in an error--either within the kernel or within the process of scoring--will count against your daily submission limit and will not return the specific error message. This is necessary to prevent probing the private test set.
Hi Sohier,
\n\nCan we train the model locally and use the kernel for infer only, just like the imet competition?
", "votes": 2, "replies": [ { "id": 563347, "author_name": "Maxwell", "author_url": "", "post_date": "2019-06-28T07:18:24.957000", "content": "I think you can.\nCheck the kernel requirements.\nhttps://www.kaggle.com/c/aptos2019-blindness-detection/overview/kernels-requirements
\n\n\n", "votes": 1, "replies": [] }, { "id": 563582, "author_name": "Yiheng Wang", "author_url": "", "post_date": "2019-06-28T13:05:01.310000", "content": "You can still train a model offline, upload it as a dataset, and use the kernel exclusively to perform inference.
\n
Thanks Maxwell
", "votes": 1, "replies": [] }, { "id": 563709, "author_name": "Sohier Dane", "author_url": "", "post_date": "2019-06-28T15:39:34.340000", "content": "Yes, Maxwell is correct.
", "votes": 0, "replies": [] }, { "id": 563924, "author_name": "Nikita Kozodoi", "author_url": "", "post_date": "2019-06-28T19:30:30.593000", "content": "Hi @sohier, just to clarify: if I train my model offline and upload it, I can keep it as a private dataset, right?
", "votes": 1, "replies": [] }, { "id": 564255, "author_name": "redshellspy", "author_url": "", "post_date": "2019-06-29T07:51:59.660000", "content": "Yes you can do that.
", "votes": 1, "replies": [] }, { "id": 565975, "author_name": "Sohier Dane", "author_url": "", "post_date": "2019-07-01T16:30:59.717000", "content": "@kozodoi yes, you can keep your trained model as a private dataset.
", "votes": 1, "replies": [] } ] }, { "id": 1345596, "author_name": "Natalia Herrera", "author_url": "", "post_date": "2021-06-11T17:27:51.863000", "content": "Hi Sohier
\nI'm currently working as a PhD student working with retinal images and I would like to know if you could publish the solution file (labels of the test images) for this competition, since it already finished.
We do not have plans to release the solution file so that we can preserve the test set for future opportunities.
", "votes": 0, "replies": [] } ] }, { "id": 1101541, "author_name": "iso1048", "author_url": "", "post_date": "2020-12-04T02:11:35.157000", "content": "Hi @sohier,
\nDo you know if the '0 - No DR' class contains images of healthy/normal retina, or could these images show other ophthalmic pathologies?
\nThank you very much.
", "votes": 0, "replies": [] }, { "id": 622110, "author_name": "Basit Riaz Sheikh", "author_url": "", "post_date": "2019-09-09T09:31:26.957000", "content": "@sohier are you going to release the public test set solution file?
", "votes": 0, "replies": [] }, { "id": 607638, "author_name": "FZHuang", "author_url": "", "post_date": "2019-08-25T16:32:18.437000", "content": "Why is it so slow when I 'Select Files to Upload'? 15MB of data took about 1h.
", "votes": 0, "replies": [] }, { "id": 580320, "author_name": "Neuron Engineer", "author_url": "", "post_date": "2019-07-20T02:11:43.343000", "content": "Hi @sohier , I have a question please.
\n\nIn this synchonous competition, since the submission is already calculated on private dataset, is it safe to delete/update some datasets which produce that score?
\n\nExample. I trained my model with dataset1 and got 0.777. After that, if I accidentally update (or delete) dataset1 (and it will affect the training process), will this affect the final score?
", "votes": 0, "replies": [] }, { "id": 569912, "author_name": "Siarhei Fedartsou", "author_url": "", "post_date": "2019-07-07T14:48:18.623000", "content": "Hi Sohier!
\n\n\n\n\nYour kernel will re-run automatically against an unseen test set
\n
How size of this \"unseen\" test set is differ relative to current test set? I think we need this information to fit to execution time limit.
", "votes": 0, "replies": [ { "id": 570086, "author_name": "Siarhei Fedartsou", "author_url": "", "post_date": "2019-07-07T18:54:28.170000", "content": "Found the answer on the Data page. It will be about 13k images.
", "votes": 0, "replies": [] } ] }, { "id": 567822, "author_name": "Xuan Cao", "author_url": "", "post_date": "2019-07-04T03:40:37.333000", "content": "Hi Sohier,
\n\nCan we use custom packages that are not pre-installed in the kernel?
", "votes": 0, "replies": [ { "id": 568919, "author_name": "Sohier Dane", "author_url": "", "post_date": "2019-07-05T16:57:32.483000", "content": "Yes, as long as those packages comply with the rest of the rules like the requirement for winning solutions to be open sourced.
", "votes": 0, "replies": [] }, { "id": 570159, "author_name": "Xuan Cao", "author_url": "", "post_date": "2019-07-07T23:09:14.477000", "content": "Thanks for the clarification.
", "votes": 0, "replies": [] } ] }, { "id": 566609, "author_name": "Aarthi", "author_url": "", "post_date": "2019-07-02T11:06:57.670000", "content": "Sounds like a good use case. Thanks for hosting this competition. @sohier
", "votes": 0, "replies": [] }, { "id": 565183, "author_name": "Carlo", "author_url": "", "post_date": "2019-06-30T14:49:22.030000", "content": "Sounds like a great use case for ML! I will try to get a team together for this!
", "votes": 0, "replies": [] }, { "id": 564506, "author_name": "gunish jha", "author_url": "", "post_date": "2019-06-29T14:16:25.663000", "content": "Sounds exciting !!, will participate in this.
", "votes": 0, "replies": [] }, { "id": 564069, "author_name": "Khoi Nguyen", "author_url": "", "post_date": "2019-06-29T00:07:31.353000", "content": "Hi @sohier, it's currently 00:06 29/06 UTC as I'm typing this, however my submission count for today hasn't been reset and I'll have to wait for 24 more hours. Is this working as intended?
", "votes": 0, "replies": [] }, { "id": 566063, "author_name": "", "author_url": "", "post_date": "2019-07-01T18:56:44.397000", "content": "", "votes": 0, "replies": [] }, { "id": 607649, "author_name": "Kiran Hegde", "author_url": "", "post_date": "2019-08-25T16:53:55.083000", "content": "Thanks
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "563162": "Welcome to the APTOS 2019 Blindness Detection challenge! It's been several years since Kaggle ran a similar competition, so we're excited to see what the progress in ML techniques can contribute to this problem. This time we have teamed up with the Aravind Eye Care Center in India which is committed to using AI to help individuals in underserved areas.\n\nWe've retained as much data as possible for the test set to ensure the leaderboard scores are meaningful, so the train set is relatively small. You may wish to take advantage of the external data rules to incorporate other diabetic retinopathy datasets into your training set. Just don't forget to declare your sources in the external data thread\n", "570261": "Hi @sohier , thanks for organizing this competition. \nAs discussed in [another thread](https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/98856#latest-570171), there are duplicated images in train and test. \n\nI wrote up a simple kernel folked from @manojprabhaakr 's original one. \nhttps://www.kaggle.com/maxwell110/duplicated-list-csv-file/output \n \nAfter I checked through the duplicated list, I confirmed some duplicated images have different labels. \n\n**examples**\n\n\n\nHow should we handle this ?\n\nThanks in advance.", "570046": "As I understood, kernels are run during submission with unseen test dataset. Yes? And public leaderboard is calculated using 15% of unseen test dataset? If it is a true, then why do we need current test dataset?", "565667": "Maybe close the previous competition's submission entrance is better.", "565650": "Thank you for hosting this competition. \nBut I and some kagglers suffered submission error. \nHere is my post. \nhttps://www.kaggle.com/c/aptos2019-blindness-detection/discussion/98058#latest-565488 \nI would like to show the error log to us when submission fail. \nThis will save our precious time:)", "563179": "Hi Sohier,\n\nCan we train the model locally and use the kernel for infer only, just like the imet competition?", "1345596": "Hi Sohier\nI'm currently working as a PhD student working with retinal images and I would like to know if you could publish the solution file (labels of the test images) for this competition, since it already finished. ", "1101541": "Hi @sohier,\n\nDo you know if the '0 - No DR' class contains images of healthy/normal retina, or could these images show other ophthalmic pathologies?\n\nThank you very much.", "622110": "@sohier are you going to release the public test set solution file? ", "607638": "Why is it so slow when I 'Select Files to Upload'? 15MB of data took about 1h.", "580320": "Hi @sohier , I have a question please.\n\nIn this synchonous competition, since the submission is already calculated on private dataset, is it safe to delete/update some datasets which produce that score?\n\nExample. I trained my model with dataset1 and got 0.777. After that, if I accidentally update (or delete) dataset1 (and it will affect the training process), will this affect the final score?", "569912": "Hi Sohier!\n\n> Your kernel will re-run automatically against an unseen test set\n\nHow size of this \"unseen\" test set is differ relative to current test set? I think we need this information to fit to execution time limit.", "567822": "Hi Sohier, \n\nCan we use custom packages that are not pre-installed in the kernel? ", "566609": "Sounds like a good use case. Thanks for hosting this competition. @sohier ", "565183": "Sounds like a great use case for ML! I will try to get a team together for this!", "564506": "Sounds exciting !!, will participate in this.", "564069": "Hi @sohier, it's currently 00:06 29/06 UTC as I'm typing this, however my submission count for today hasn't been reset and I'll have to wait for 24 more hours. Is this working as intended?", "566063": "", "607649": "Thanks" } }