Pre-trained models
收藏
As stated here " In addition to the dataset provided by the organizer, all contestantsare not allowed to use the fundus annotated data from any other channels (including but not limited to public fundus datasets, self-annotated datasets, etc.)", the contestants are not allowed to use any public datasets, however, what about the pre-trained models? Can the contestants use any kind of pretrained models?
Thanks in advance.
0
收藏
请登录后评论
Pre-trained models without using fundus dataset are available, such as ResNet family pretrained by ImageNet.
The rules of the competition are not allowed to use additional fundus datasets, such as REFUGE, IDRiD, etc.
What about unlabeled fundus images?
The fundus images and OCT images of the same number are paired. For Task 1 glaucoma grading, these two modalities' data share the same label. For Task 2 and Task 3, we aim to predict the coordinate value of fovea in the 2D fundus image and segment the optic disc and cup region in 2D fundus images. Hence, the labels of Task 2 and Task 3 are for the 2D fundus images.