Generate different types of random avatars.

Python3

pytorch>=1.6

cv2>=3.4

tqdm

Please download the pretrained models from:

https://drive.google.com/drive/folders/1J_99la18a2ZeIBEPQ2_xqFa_uOxkrZVI?usp=sharing

and put the files under the "models" folder. The filetree should look like:

models
  |-avatar_02_ae_07_decoder.pth
  |-avatar_02_ae_07.npy
  |-avatar_04_cartoon_face_02_decoder.pth
  |-avatar_04_cartoon_face_02.npy

Run the following line to generate png files, where 10 represents 10 output avatars, and 96 represents the size of images are 96*96.

python generate_githublike.py ./examples/avatars_githublike/ 10 96

These avators are generated by the decoder of a pretrained variational auto-encoder model. Run the following line to generate png files, where 10 represents 10 output avatars, and 96 represents the size of images are 96*96.

python generate_athletes.py ./examples/avatars_athletes/ 10 96

These avators are generated by the generator of a pretrained generative adversarial network (GAN) model. Run the following line to generate png files, where 10 represents 10 output avatars, and 96 represents the size of images are 96*96.

python generate_anime.py ./examples/avatars_anime/ 10 96

These avators are generated by an adjusted version of Stylegan (ref: https://github.com/NVlabs/stylegan).