Pretraining Image Encoders without Reconstruction via Feature Prediction Loss

Re-assign community