Visual-Semantic Refinement Network: Towards Exploring the Capabilities of Decoder in Scene Text Recognition | IEEE Conference Publication | IEEE Xplore