Static-Dynamic Interaction Networks for Offline Signature Verification

• DOI: 10.1609/aaai.v35i3.16284
• Publication Date: 2022-09-08T18:13:42Z
• Authors: Huan Li; Ping Wei; Ping Hu
• Keywords: Signature (topology); Representation; Feature (linguistics)

Offline signature verification is a challenging issue that widely used in various fields. Previous approaches model this task as static feature matching or distance metric problem of two images. In paper, we propose novel Static-Dynamic Interaction Network (SDINet) which introduces sequential representation into A image converted to sequences by assuming pseudo dynamic processes the image. extracting deep features from images describes global information signatures. with LSTM networks characterizes local dynamic-to-static attention learned refine features. Through static-to-dynamic conversion and attention, are unified compact framework. The proposed method was evaluated on four popular datasets different languages. extensive experimental results manifest strength our model.