To further improve the accuracy of multilingual off-line handwritten signature verification, this paper studies the off-line handwritten signature verification of monolingual and multilingual mixture and proposes an improved verification network (IDN), which adopts user-independent (WI) handwritten signature verification, to determine the true signature or false signature. The IDN model contains four neural network streams with shared weights, of which two receiving the original signature images are the discriminative streams, and the other two streams are the reverse stream of the gray inversion image. The enhanced spatial attention models connect the discriminative streams and reverse flow to realize message propagation. The IDN model uses the channel attention mechanism (SE) and the improved spatial attention module (ESA) to propose the effective feature information of signature verification. Since there is no suitable multilingual signature data set, this paper collects two language data sets (Chinese and Uyghur), including 100,000 signatures of 200 people. Our method is tested on the self-built data set and the public data sets of Bengali (BHsig-B) and Hindi (BHsig-H). The method proposed in this paper has the highest discrimination rate of FRR of 10.5%, FAR of 2.06%, and ACC of 96.33% for the mixture of two languages.

Target tracking algorithms based on deep learning have achieved good results in public datasets. Among them, the network tracking algorithm based on Siamese tracking has a high accuracy and fast speed, which has attracted significant attention. However, the Siamese tracker uses the AlexNet network as its backbone and the network layers are relatively shallow, so it does not make full use of the ability of the deep neural network. If only the backbones of target tracking are replaced, there will be no obvious improvement, such as in the cases of ResNet and Inception. Therefore, this paper designs a wider and deeper network structure. At a wider level, a mechanism that can adaptively adjust the receptive field (RF) size is designed. Firstly, multiple branches are divided by the split operator, and each branch has a different size of kernel corresponding to a different size of RF; then, the fuse operator is used to fuse the information of each branch to obtain the selection weights; and finally, according to the selection, the aggregation feature map is weighted. At a deeper level, a new kind of residual models is designed. The channel is simplified by pruning in order to improve the tracking speed. According to the above, a wider and deeper Siamese network was proposed in this paper. The experimental results show that the structure proposed in this paper achieves a good tracking effect and real-time performance on six kinds of datasets. The proposed tracker achieves an SUC and Prec of LaSOT of 0.569 and 0.571, respectively.

No abstract

No abstract

No abstract