We introduce a novel dataset of Subjective Assessments of Legibility in Ancient Manuscript Images (SALAMI) to serve as a ground truth for the development of quantitative evaluation metrics in the field of digital text restoration.
This dataset consists of 250 images of 50 manuscript regions with corresponding spatial maps of mean legibility and uncertainty, which are based on a study conducted with 20 experts of philology and paleography.
Description of files:
- input - rated test images
- mean_score_maps - spatial maps of mean legibility
- std_maps - spatial maps of uncertainty (standard deviation of legibility)
- images.json - definition of source images contained in the dataset
- users.json - list of participants with their respective properties
- assessments.json - the main data generated by our experiments.
- salami_proc.py - contains python functions to process the .json files named above
- salami_proc_usage.py - uses the functions from salami_proc.py to reproduce the output images and statistical results described in the accompanying paper
- salami_llm.R - documents the linear mixed models analysis performed in R