Mixed Supervised Object Detection by Transferring Mask Prior and Semantic Similarity

Object detection has achieved promising success, but requires large-scale fully-annotated data, which is time-consuming and labor-extensive.