Closed SalAlba closed 5 years ago
This is a good question. For the method used in this repository, you need to go through the long process each time.
If you instead use a better algorithm, you need to only train once per style. Then the actual act of applying the style to any given input image is super fast. Please see e.g. https://github.com/lengstrom/fast-style-transfer.
Good project and explanation, I have one question when we train our model on specific style, can we reuse it with another images ? with less time of moving style ?
Do I have make this long process every time whenever I want one style to another image ?