Trans-Cycle: Unpaired Image-to-Image Translation Network by Transformer

Transformer recently has made remarkable progress in various computer vision tasks. In this paper, we design a novel transformer-based framework for image-to-image translation under the condition of limited computing resources. We conducted a comparative analysis between the model introduced in this paper and mainstream methodologies prevalent in the field, employing various publicly available datasets. The evaluation encompassed the assessment of numerical metrics and the examination of visualization outcomes. Compared with the recent unpaired translation network, our model has a more powerful deformation ability and a more refined distinction between image sub-regions. It has better performance on multiple datasets and much development potential.