The Second Place Solution for The 4th Large-scale Video Object Segmentation Challenge-Track 3: Referring Video Object Segmentation