Open-source datasets, code and pretrained models.
SIW-10 dataset from the ICDAR 2015 paper "Automatic Script Identification in the Wild". A benchmark for script identification in natural scene images.
