DropMAE: Masked Autoencoders with Spatial-Attention Dropout for Tracking Tasks | IEEE Conference Publication | IEEE Xplore