Closed songxiaoguo closed 1 year ago
It depends on the probability in all_partition
. If all datasets have a probability larger than 0, then they will participate in the training. The train
only indicates the mode and will influence the augmentation, etc.
Sorry to bother the author, but may I ask if 'train' is set to true in JointDatasetCam, which means that all data will participate in the training?