Identifying Cross-Depicted Historical Motifs

Abstract

Cross-depiction is the problem of identifying the same object even when it is depicted in a variety of manners. This is a common problem in handwritten historical documents image analysis, for instance when the same letter or motif is depicted in several different ways. It is a simple task for humans yet conventional heuristic computer vision methods struggle to cope with it. In this paper we address this problem using stateof-the-art deep learning techniques on a dataset of historical watermarks containing images created with different methods of reproduction, such as hand tracing, rubbing, and radiography.To study the robustness of deep learning based approaches to the cross-depiction problem, we measure their performance on two different tasks: classification and similarity rankings. For the former we achieve a classification accuracy of 96 % using deep convolutional neural networks. For the latter we have a false positive rate at 95% true positive rate of 0.11. These results outperform state-of-the-art methods by a significant margin.

A Comprehensive Study of ImageNet Pre-Training for Historical Document Image Analysis

“…When it comes to cross-domain transfer learning, its usefulness -especially from natural images such as ImageNet -is subject of an open discussion. There are cases where transfer learning across domains has been proven to be successful [17], [18]. In contrast, there is literature suggesting that this technique is harmful to the final performance of the networks.…”

Combining graph edit distance and triplet networks for offline signature verification

“…In this work we are considering the classification pretraining introduced by Pondenkandath et al (2018). In this classification pretraining step, the neural network is trained for classification with cross-entropy loss on the same training set before training for similarity.…”

“…Furthermore, we investigate an additional pretraining step for the neural networks in which we train for classification before training for similarity. This pretraining step has been shown to increase the network performances especially when for each class, there is little amount of labeled data available (Pondenkandath et al, 2018). Finally, we are using two more test sets and more evaluation metrics to compare our framework against more published results in our experimental evaluation.…”

DeepDIVA: A Highly-Functional Python Framework for Reproducible Experiments

“…Note that more details on these experiments and how to tackle the cross-depiction problem can be found explained in detail in the original work [22].…”