Snowballing framework between the oil slick dataset and the deep learning model. (IMAGE)
Caption
Snowballing framework between the oil slick dataset and the deep learning model. The left panel (in red) represents the process of data preparation, and the right panel (in blue) represents the training, evaluation, and prediction of the model. We divided samples in the dataset into three classes (dashed boxes in the middle): positive samples with qualified masks can be directly used for the training; positive samples with unqualified masks need to be refined by the model; and the diversity of negative samples need to be augmented by the model.
Credit
Journal of Remote Sensing
Usage Restrictions
Credit must be given to the creator.
License
CC BY