![\"CorrectedLB\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F876618%2Fb1efad1cf7c456c8f10ee182ae37ee4f%2Ffinal_LB_corrected.jpg?generation=1659319235706154&alt=media\")
Thank you all for participating in this competition.
\nWe are thankful for your time and effort to create the best models to classify body parts in X-rays. There were more than 1200 submissions! 113 teams and 122 participants! Wow!
\nWe hope your models will allow more seamless integration of other machine learning models in the radiology workflow.
\nSome of you identified errors in the labels. We understand that labeling is an art and part of the ML development process. To guarantee a fair game, we decided to keep the test set labels as provided originally at the beginning of the competition. The final private LB reflects those labels.
\nJust as a matter of curiosity, we did our best to correct the labels in the test set, having 4 radiologists annotating each image, and rerun the calculation of the entire LB offline so you know how your models would behave with a more correct set of labels. These results will be disclosed in this post as soon as we finish calculating them. We will not change the final LB.
\nWe hope you all had fun and learned new skills by participating in this challenge.
\nWe thank all the annotators who made this work possible.
\nWe invite you to participate in the UNIFESP Chest CT Fatty Liver Competition
\n@felipekitamura and @eduardofarina
", "messages": [ { "id": "1879246", "postDate": "08/01/2022 00:11:41", "content": "Thank you all for participating in this competition.
\nWe are thankful for your time and effort to create the best models to classify body parts in X-rays. There were more than 1200 submissions! 113 teams and 122 participants! Wow!
\nWe hope your models will allow more seamless integration of other machine learning models in the radiology workflow.
\nSome of you identified errors in the labels. We understand that labeling is an art and part of the ML development process. To guarantee a fair game, we decided to keep the test set labels as provided originally at the beginning of the competition. The final private LB reflects those labels.
\nJust as a matter of curiosity, we did our best to correct the labels in the test set, having 4 radiologists annotating each image, and rerun the calculation of the entire LB offline so you know how your models would behave with a more correct set of labels. These results will be disclosed in this post as soon as we finish calculating them. We will not change the final LB.
\nWe hope you all had fun and learned new skills by participating in this challenge.
\nWe thank all the annotators who made this work possible.
\nWe invite you to participate in the UNIFESP Chest CT Fatty Liver Competition
\n@felipekitamura and @eduardofarina
", "rawMarkdown": "Thank you all for participating in this competition.\n\nWe are thankful for your time and effort to create the best models to classify body parts in X-rays. There were more than 1200 submissions! 113 teams and 122 participants! Wow!\n\nWe hope your models will allow more seamless integration of other machine learning models in the radiology workflow.\n\nSome of you identified errors in the labels. We understand that labeling is an art and part of the ML development process. To guarantee a fair game, we decided to keep the test set labels as provided originally at the beginning of the competition. The final private LB reflects those labels. \n\nJust as a matter of curiosity, we did our best to correct the labels in the test set, having 4 radiologists annotating each image, and rerun the calculation of the entire LB offline so you know how your models would behave with a more correct set of labels. These results will be disclosed in this post as soon as we finish calculating them. We will not change the final LB.\n\nWe hope you all had fun and learned new skills by participating in this challenge.\n\nWe thank all the annotators who made this work possible.\n\nWe invite you to participate in the [UNIFESP Chest CT Fatty Liver Competition](https://www.kaggle.com/competitions/unifesp-fatty-liver)\n\n@felipekitamura and @eduardofarina", "votes": null }, { "id": "1879314", "postDate": "08/01/2022 02:07:43", "content": "The LB below has three differences from the official LB:
\nThis LB was calculated out of curiosity, so you know how well your model would perform in the entire test set with corrected labels. Congratulations to everyone, regardless of your position!
\n
@felipekitamura
\nThank for organising the competition!
I made a poor choice of subs and shaked down, but I learnt a lot from the competition.
\nAnd thanks for the additional LBs, I'm glad we were able to make a good model.
Personally, I'm going to take part in the RSNA competition, so thank you again for hosting!😄
", "rawMarkdown": "felipekitamura \nThank for organising the competition!\n\nI made a poor choice of subs and shaked down, but I learnt a lot from the competition.\nAnd thanks for the additional LBs, I'm glad we were able to make a good model.\n\nPersonally, I'm going to take part in the RSNA competition, so thank you again for hosting!😄", "votes": null }, { "id": "1879806", "postDate": "08/01/2022 09:21:11", "content": "Thanks for joining the competition and putting a lot of effort on doing a great job! Congratulations!
", "rawMarkdown": "Thanks for joining the competition and putting a lot of effort on doing a great job! Congratulations!", "votes": null } ], "comments": [ { "id": 1879314, "author_name": "felipekitamura", "author_url": "", "post_date": "08/01/2022 02:07:43", "content": "The LB below has three differences from the official LB:
\nThis LB was calculated out of curiosity, so you know how well your model would perform in the entire test set with corrected labels. Congratulations to everyone, regardless of your position!
\n
@felipekitamura
\nThank for organising the competition!
I made a poor choice of subs and shaked down, but I learnt a lot from the competition.
\nAnd thanks for the additional LBs, I'm glad we were able to make a good model.
Personally, I'm going to take part in the RSNA competition, so thank you again for hosting!😄
", "votes": null, "replies": [ { "id": 1879806, "author_name": "eduardofarina", "author_url": "", "post_date": "08/01/2022 09:21:11", "content": "Thanks for joining the competition and putting a lot of effort on doing a great job! Congratulations!
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1879246": "Thank you all for participating in this competition.\n\nWe are thankful for your time and effort to create the best models to classify body parts in X-rays. There were more than 1200 submissions! 113 teams and 122 participants! Wow!\n\nWe hope your models will allow more seamless integration of other machine learning models in the radiology workflow.\n\nSome of you identified errors in the labels. We understand that labeling is an art and part of the ML development process. To guarantee a fair game, we decided to keep the test set labels as provided originally at the beginning of the competition. The final private LB reflects those labels. \n\nJust as a matter of curiosity, we did our best to correct the labels in the test set, having 4 radiologists annotating each image, and rerun the calculation of the entire LB offline so you know how your models would behave with a more correct set of labels. These results will be disclosed in this post as soon as we finish calculating them. We will not change the final LB.\n\nWe hope you all had fun and learned new skills by participating in this challenge.\n\nWe thank all the annotators who made this work possible.\n\nWe invite you to participate in the [UNIFESP Chest CT Fatty Liver Competition](https://www.kaggle.com/competitions/unifesp-fatty-liver)\n\n@felipekitamura and @eduardofarina", "1879314": "The LB below has three differences from the official LB:\n1. It was calculated in the entire test (public + private)\n2. It considered the corrected labels (4 annotators + adjudication)\n3. It was calculated based on the best submission each participant made, which often does not match the submission chosen by the participant to count in the official LB.\n\nThis LB was calculated out of curiosity, so you know how well your model would perform in the entire test set with corrected labels. Congratulations to everyone, regardless of your position!\n\n", "1879349": "felipekitamura \nThank for organising the competition!\n\nI made a poor choice of subs and shaked down, but I learnt a lot from the competition.\nAnd thanks for the additional LBs, I'm glad we were able to make a good model.\n\nPersonally, I'm going to take part in the RSNA competition, so thank you again for hosting!😄", "1879806": "Thanks for joining the competition and putting a lot of effort on doing a great job! Congratulations!" }, "source": "meta" }