FLSL: Feature-level Self-supervised Learning

Neural Information Processing Systems 

MOCOv3) target primarily on representations at instance level and do not generalize well to dense prediction tasks, such as object detection and segmentation.