AGUnet: Annotation-guided U-net for fast one-shot video object segmentation
作者：
Highlights：
• An annotation guided U-net is proposed to achieved fast video object segmentation.
• The annotation information obtained by Siamese networks is incorporated into a U-net.
• AGUnet can be trained in an end-to-end manner on datasets with only static images.
• AGUnet is target-focusing and can quickly adapt the trained model to segment targets.
• Experiments demonstrate the effectiveness and fast speed of the proposed AGUnet.
摘要
•An annotation guided U-net is proposed to achieved fast video object segmentation.•The annotation information obtained by Siamese networks is incorporated into a U-net.•AGUnet can be trained in an end-to-end manner on datasets with only static images.•AGUnet is target-focusing and can quickly adapt the trained model to segment targets.•Experiments demonstrate the effectiveness and fast speed of the proposed AGUnet.
论文关键词：Fully-convolutional Siamese network,U-net,Interactive image segmentation,Video object segmentation
论文评审过程：Received 16 June 2019, Revised 29 June 2020, Accepted 7 August 2020, Available online 13 August 2020, Version of Record 11 September 2020.
论文官网地址：https://doi.org/10.1016/j.patcog.2020.107580
