Semi-supervised Feature Learning For Improving Writer Identification
This work addresses writer identification, a domain-specific task in document analysis, with an incremental improvement over existing methods.
The paper tackled the problem of offline writer identification by proposing a semi-supervised feature learning pipeline that uses weighted label smoothing regularization to train with both labeled and unlabeled data, resulting in significant improvements over the baseline and competitive performance on benchmark datasets like ICDAR2013 and CVL.
Data augmentation is usually used by supervised learning approaches for offline writer identification, but such approaches require extra training data and potentially lead to overfitting errors. In this study, a semi-supervised feature learning pipeline was proposed to improve the performance of writer identification by training with extra unlabeled data and the original labeled data simultaneously. Specifically, we proposed a weighted label smoothing regularization (WLSR) method for data augmentation, which assigned the weighted uniform label distribution to the extra unlabeled data. The WLSR method could regularize the convolutional neural network (CNN) baseline to allow more discriminative features to be learned to represent the properties of different writing styles. The experimental results on well-known benchmark datasets (ICDAR2013 and CVL) showed that our proposed semi-supervised feature learning approach could significantly improve the baseline measurement and perform competitively with existing writer identification approaches. Our findings provide new insights into offline write identification.