The paper states that "The domain translation functions (i.e., $g{\phi^A}^A$ and $g{\phi^B}^B$) are no longer used in inference", but no reasonable explanation is provided. I have doubts as to whether the domain translation functions has been properly trained to achieve the expected goal. If you have trained the model, would you be willing to visualize the prediction results directly using the domain translation functions? I would greatly appreciate it!
The paper states that "The domain translation functions (i.e., $g{\phi^A}^A$ and $g{\phi^B}^B$) are no longer used in inference", but no reasonable explanation is provided. I have doubts as to whether the domain translation functions has been properly trained to achieve the expected goal. If you have trained the model, would you be willing to visualize the prediction results directly using the domain translation functions? I would greatly appreciate it!