CycleMatch: A cycle-consistent embedding network for image-text matching

作者：

Highlights：

• A novel deep cycle-consistent embedding network (CycleMatch) for image-text matching.

• Maintain both inter-modal correlations and intra-modal consistency.

• Late-fusion approaches to efficiently integrate the matching scores of multiple embedding features.

• Competitive with the state-of-the-art performance on two well-known multimodal benchmarks.

摘要

•A novel deep cycle-consistent embedding network (CycleMatch) for image-text matching.•Maintain both inter-modal correlations and intra-modal consistency.•Late-fusion approaches to efficiently integrate the matching scores of multiple embedding features.•Competitive with the state-of-the-art performance on two well-known multimodal benchmarks.

论文关键词：Image-text matching,Embedding,Deep neural networks,Late-fusion inference

论文评审过程：Received 20 September 2018, Revised 15 March 2019, Accepted 1 May 2019, Available online 4 May 2019, Version of Record 8 May 2019.

论文官网地址：https://doi.org/10.1016/j.patcog.2019.05.008