mapkurator-spotter

This repository contains the implementations of the text spotter for scanned historical maps, spotter-v2, which has been deployed in the mapKurator system. Spotter-v2 is built upon Deformable-DETR and TESTR.

Please refer to the mapKurator documentation for details.

Model weights

All the model weights can be found here

Model weight for Spotting English Text

You can donwload the model weight and a model card. The configuration file for inferencing is under ./configs.

Model weight for Spotting Coordinates

You can donwload the model weight. The configuration file for inferencing is under ./configs.

Model weights for Multilingual Text Spotting

mapkurator-spotter supports multilingual text spotting. Please add the corresponding configuration file and weight to run the model.

Supported Language	Config folder	Model weight	Model card
Russian	config	weight	model card
Chinese + Japanese	config	weight	model card
Arabic	config	weight	model card

Acknowledgement

Thanks to AdelaiDet for a standardized training and inference framework, and Deformable-DETR for the implementation of multi-scale deformable cross-attention, and  TESTR for the implementation of Deformable-DETR text spotter for scene images.