Contrastive Masked Autoencoders are Stronger Vision Learners

Proposes CMAE, a new vision SSL pre-training, which achieves the SotA performance on classification, semantic segmentation and object detection.

This paper presents a new self-supervised pre-training method for vision representations called CMAE. It achieves state-of-the-art performance on various vision tasks such as image classification, semantic segmentation, and object detection. Business operations that involve computer vision could benefit from this method by improving their models' performance and capabilities.