Machine learning of spatio-temporal manifolds for source-free video domain adaptation

A system for training a model, comprising:

• a hardware processor;
• and
• a memory that stores a computer program which, when executed by the hardware processor, causes the hardware processor to:
• perform spatial augmentation on an unlabeled input video to generate spatially augmented video;
• perform temporal augmentation on the input video to generate temporally augmented video;
• generate predictions, using a model that was pre-trained on a labeled dataset, for the unlabeled input video, the spatially augmented video, and the temporally augmented video;
• and
• adapt parameters of the model using the predictions while enforcing temporal consistency, spatial consistency, and historical consistency, including an information maximization loss that has a first term that minimizes an entropy of probability and a second term that maximizes an entropy of a mean of predictions over videos in a target domain.