Office Documents Classification under Limited Sample
Paweł Baranowski, Adrian Stepniak · Journal of information and organizational sciences · 2022
Deep convolutional neural networks (CNNs) became an industry standard in imageprocessing. However, in order to keep their high efficiency, a large annotated sampleis required in the case of supervised learning. In this paper we apply the techniquesspecific for relatively small sample to a court files dataset. Specifically, we proposetransfer learning and semisupervised learning to classify scanned page as having a tableor not. We use four CNNs architectures established in the literature and find thattransfer learning improves the classification performance, compared to the fullysupervised learning. This result is especially evident in the scenarios where only a partof convolutioanl layers are transferred. The gains from semisupervised learning areambiguous, as the results vary over CNNs architectures. Overall, our results show thatoffice documents classification can achieve high accuracy when transferring initialconvolutional layers is applied.