encoder4editing with MobileNetV3

The demo of different styles images of e4e-mbv3 framework.

Description

Re-implementation of e4e that use mobilenet-v3 and stylegan2-1024p. Then mix different styles by stylegan2 and convert to openvino.

Citation

This code is heavily based on encoder4editing and stylegan2-pytorch. Thanks omertov and rosinality so much to make his work available 🙏🙏🙏

Pretrained Models

链接: https://pan.baidu.com/s/1JlJTyBmFLZhju4MpDHLdAA 密码: 086q

Google Drive

Training Log

Tensorboard

The loss of training.

Test

python scripts/inference.py \
--images_dir=./test_images/ \
--save_dir=./sample \
./ckpt/best_model.pt

More

Inference on openvino and more demo