Abstract | ||
---|---|---|
Identification of scribes from historical manuscripts has remained an equally interesting problem for paleographers as well as the pattern classification researchers. Though significant research endeavors have been made to address the writer identification problem in contemporary handwriting, the problem remains challenging when it comes to historical manuscripts primarily due to the degradation of documents over time. This study targets scribe identification from ancient documents using Greek handwriting on the papyri as a case study. The technique relies on segmenting the handwriting from background and extracting keypoints which are likely to carry writer-specific information. Using the handwriting keypoints as centers, small fragments (patches) are extracted from the image and are employed as units of feature extraction and subsequent classification. Decisions from fragments of an image are then combined to produce image-level decisions using a majority vote. Features are learned using a two-step fine-tuning of convolutional neural networks where the models are first tuned on contemporary handwriting images (relatively larger dataset) and later tuned to the small set of writing samples under study. The preliminary findings of the experimental study are promising and establish the potential of the proposed ideas in characterizing writer from a challenging set of writing samples. |
Year | DOI | Venue |
---|---|---|
2021 | 10.1007/978-3-030-86198-8_32 | DOCUMENT ANALYSIS AND RECOGNITION, ICDAR 2021 WORKSHOPS, PT I |
Keywords | DocType | Volume |
Scribe identification, FAST keypoints, ConvNets, IAM Dataset, Greek Papyrus | Conference | 12916 |
ISSN | Citations | PageRank |
0302-9743 | 0 | 0.34 |
References | Authors | |
0 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Sidra Nasir | 1 | 0 | 0.34 |
Imran Siddiqi | 2 | 421 | 36.56 |
Momina Moetesum | 3 | 1 | 1.70 |