Open johndzengpft opened 1 year ago
First, ensure stylized images fed to CCN are high-quality and style-faithfulness as possible, manual picking may help. Second, more steps may need for sufficient training, we use 30w steps on average.
@menyifang Excuse me, I want to ask you questions:
4.The training was interrupted, and the resume_epoch in the configuration file was modified to start from 42999. The loss training 20k has not changed. Is this normal?
Hi, I have trained the design style by the following steps with default config:
python generate_data.py --style design
python extract_align_faces.py
and pick 200 style imagespython train_localtoon.py
with total 19962 source images & 79832 style imagesThe result of 10,000 step is not as good as official pre-train:![9999_face_result](https://user-images.githubusercontent.com/128478543/228206622-c220a489-dd2b-4ebb-8c0f-b4b64f2e49a6.jpg)
Did I do any step wrong? Thank you very much!