The included repositories contain code and model files for tools developed as part of the the DALAI project (September 2021 - August 2023). The project was funded by the European Regional Development Fund’s ”Sustainable growth and jobs 2012–2020” programme and the City of Mikkeli.
The common aim of the different tools is to facilitate the automation of the digitisation and description of cultural heritage materials which are in the holdings of archives and other memory organisations.
Click here for more information on the included repositories
Repository | Domain | Content |
---|---|---|
CornerAPI | Image Classification | Code for an API that detects torn corners and edges from document images. |
EmptyAPI | Image Classification | Code for an API that detects empty pages from document images. |
PostitAPI | Image Classification | Code for an API that detects post-it/sticky notes from document images. |
WritingtypeAPI | Image Classification | Code for an API that classifies document images based on the writing types(s) (handwritten, typewritten, combination) they contain. |
FaultyImageAPI | Image Classification | Code for an API that combines the classification models listed above. |
NER_API | Named Entity Recognition | Code for an API that performs named entity recognition from text input in Finnish. |
Train_BERT_NER | Named Entity Recognition | Code for training Finnish named entity recognition (NER) model based on BERT language model. |
Empty_training | Image Classification | Code for training a neural network model to detect empty pages from document images. |
Train_document_classification | Image Classification | Code for training a neural network model to classify input documents into distinct classes based on the type/format of the document. |
Train_fault_detection | Image Classification | Code for training a neural network model to detect faults like folded corners or sticky notes from document images. |
Train_writing_type | Image Classification | Code for training a neural network model to classify document images based on the writing types(s) (handwritten, typewritten, combination) they contain. |
Table_segmentation | Image Segmentation | Code for segmenting table structures and detecting text content in document images. |
Some of the tools are also available via Arkkiivi web user interface.