The difference between the updated training data and the previous training data

The difference between the updated training data and the previous training data  

  By: JL_Ng on Sept. 28, 2024, 10:55 a.m.

Thanks for organizing such an excellent challenge, I utilized the updated training data to train my model. But I think there is a significant difference between the updated training data and the previous training data, particularly in zones 4 and 5. I'm curious about if you have updated all the labels in your entire dataset, including the test sets? If the entire dataset has been updated, I think it would be more appropriate to use the updated training data.

And I also noticed that some aortic branches in the two versions of the training data were annotated with different lengths. None of us want the inconsistency in the annotation process to affect the evaluation of our algorithms. I would like to share several pictures that highlight the differences between the previous training data and the updated training data on google drive. "https://drive.google.com/drive/folders/1QacgAT43PzdtSVY9nEf3Jlfc7J3xXksX?usp=drive_link"

 Last edited by: JL_Ng on Sept. 28, 2024, 11:17 a.m., edited 1 time in total.

Re: The difference between the updated training data and the previous training data  

  By: JL_Ng on Sept. 28, 2024, 10:57 a.m.

Unfortunately, I only have 2 chances for test phase submission, so I would like to confirm whether the annotation standards for the current test data are consistent with those of the training data. Additionally, would there be a significant difference in performance on the test data if different versions of the training data(original VS updated training data) were used for training?