tohinz/ConSinGAN
PyTorch implementation of "Improved Techniques for Training Single-Image GANs" (WACV-21)
Implements progressive multi-resolution training where the generator capacity and training image resolution increase iteratively, with selective layer-wise learning rate scaling to control fidelity-diversity tradeoffs. Supports multiple tasks beyond generation—image animation, harmonization, and editing—each trainable on single images in 5-25 minutes on a single GPU. Provides fine-tuning capabilities and arbitrary-size generation modes, with Tensorboard logging for real-time training monitoring.
443 stars. No commits in the last 6 months.
Stars
443
Forks
77
Language
Python
License
MIT
Category
Last pushed
Jan 13, 2022
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/tohinz/ConSinGAN"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
yunjey/domain-transfer-network
TensorFlow Implementation of Unsupervised Cross-Domain Image Generation
taesungp/contrastive-unpaired-translation
Contrastive unpaired image-to-image translation, faster and lighter training than cyclegan (ECCV...
PaddlePaddle/PaddleGAN
PaddlePaddle GAN library, including lots of interesting applications like First-Order motion...
sagiebenaim/DistanceGAN
Pytorch implementation of "One-Sided Unsupervised Domain Mapping" NIPS 2017
ZPdesu/SEAN
SEAN: Image Synthesis with Semantic Region-Adaptive Normalization (CVPR 2020, Oral)