Large-Scale Historical Watermark Recognition: dataset and a new consistency-based approach
This work addresses a practical challenge for archivists and historians by enabling large-scale, fine-grained recognition of historical watermarks, though it is incremental as it builds on standard deep learning methods with specific improvements.
The paper tackles the problem of historical watermark recognition by introducing a large public dataset with over 6,000 new photographs and a consistency-based approach using mid-level deep features and spatial filtering, achieving 55% top-1 accuracy on a 16,753-class one-shot cross-domain recognition task.
Historical watermark recognition is a highly practical, yet unsolved challenge for archivists and historians. With a large number of well-defined classes, cluttered and noisy samples, different types of representations, both subtle differences between classes and high intra-class variation, historical watermarks are also challenging for pattern recognition. In this paper, overcoming the difficulty of data collection, we present a large public dataset with more than 6k new photographs, allowing for the first time to tackle at scale the scenarios of practical interest for scholars: one-shot instance recognition and cross-domain one-shot instance recognition amongst more than 16k fine-grained classes. We demonstrate that this new dataset is large enough to train modern deep learning approaches, and show that standard methods can be improved considerably by using mid-level deep features. More precisely, we design both a matching score and a feature fine-tuning strategy based on filtering local matches using spatial consistency. This consistency-based approach provides important performance boost compared to strong baselines. Our model achieves 55% top-1 accuracy on our very challenging 16,753-class one-shot cross-domain recognition task, each class described by a single drawing from the classic Briquet catalog. In addition to watermark classification, we show our approach provides promising results on fine-grained sketch-based image retrieval.