Yuxiang Wei

Hi, I encounter the same problem, have you found the reason?

> Has anyone achieved good results when projecting custom images? I used encoders suggested by @PaulCzaban and results are bad. But I noticed that if I set `--num-steps` to a...

Hi, If you just want to get the frontalized faces without calculate the metrics, you can first modify the code of `line 50-72` of `test.py` as follow: ```python if opt.datamode...

The frontalized images are saved by the visualizer.display_test_results(). You can modify the `test_dir` attribute in `line 76` in [Visualizer.py](https://github.com/csyxwei/FFWM/blob/d42c578cabe1b81c6b1bb0c3cb707b190fca3c68/util/visualizer.py#L76) to specific the save path. If default, it will be saved...

FFWM can not convert a front face into a face at given angle.

Hi, could you give some examples, e.g. original images and processed images ?

### Prepare your dataset You should refer to the `README` in data_process to prepare your dataset, and organise your data as follow: ``` dataset └── your_dataset_name └── images ``` ###...

Hi, could you provide your test results? FFWM is trained on MultiPIE dataset, which is collected in controlled setting, while CelebA is a chanllage in-the-wild dataset. So it may not...

I also cannot run the project on 32G V100 when I create the environment with [environment.yaml](https://github.com/XavierXiao/Dreambooth-Stable-Diffusion/blob/main/environment.yaml), which uses pytroch1.10.2. However, when I update the pytorch to 1.12.1, I can run...

Nice trick, it reduces the memory from 31G to 27G.