AL-RSSD (Document Attribute Localization – Russian Semi-Structured Documents)

Description: The dataset contains Russian semi-structured documents of three types: Form “OS15” (18 images) “Invoice 1c” document (400 images) “Consignment note” (300 images) The total number of images is 718. Form “OS15” is presented with one template, “Invoice 1c” – with two different templates and “Consignment note” – with three templates. Dataset Structure: Folder "images" contains sorted in three folders (concerning its type) document images. Folder "ground_truth" contains sorted in three folders (concerning its type) JSON files. Each JSON file corresponds to each document image, contains its type and the text values for each attribute. Folder "supplementary_materials" images with highlighted target attributes are presented Files with an extension "<subset_name>.lst" list the files associated with the subset