Binary Inscription Character Inpainting Based on Improved Context Encoders

Due to the limitations of textual datasets, there are currently few studies on character inpainting. Second, unlike image inpainting, inpainting cannot replicate results and some methods even inaccessible. Finally, extraction features in existing model is insufficient, resulting being unable reconstruct characters based complete accurate features. In this paper, we first threshold obtained inscription dataset obtain a visually better-binarized dataset. improve Context encoders design BCEs(Binary Encoders) add dilated convolution learn structural character. It has been experimentally proven that BCEs not only slightly better than same field, but can also restore real-life with missing strokes.