Analysis of Text Image Super Resolution Loss Function Based on OPCA
This paper describes a loss function based on OPCA,which does not require learnable parameters or additional prior knowledge,and can serve as a useful supplement to existing loss functions to enhance the model's ability to represent high-dimensional features.The experimental results demonstrate that the loss function based on OPCA generally improves the text recognition accuracy of multiple text super-resolution models,and this improvement does not increase the inference time of the models.
scene text recognitionsuper-resolutionOptimization Principal Component Analysis(OPCA)loss function
NETL
NSTL
万方数据
