Skip to content

Latest commit

 

History

History
47 lines (28 loc) · 4.98 KB

MODELZOO.md

File metadata and controls

47 lines (28 loc) · 4.98 KB

imaginaire_logo.svg

Imaginaire

Model Zoo

Introduction

We provide a wide range of pretrained imaginaire models for different tasks. All the models were trained using an NVIDIA DGX 1 machine with 8 32GB V100 using NVIDIA PyTorch docker 20.06 or later.

Algorithm Name Task Model Resolution
pix2pixHD Cityscapes, segmentation to image download 1024x512
SPADE COCO-Stuff, segmentation to image download 256x256

Unsupervised Image-to-Image Translation

Algorithm Name Task Model Resolution
UNIT Winter <-> Summer download 256x256
MUNIT AFHQ Dog <-> Cat download 256x256
FUNIT AniamlFaces download 256x256
COCO-FUNIT AniamlFaces download 256x256
COCO-FUNIT Mammals, Full body animal translation download 256x256

Video-to-video Translation

Algorithm Name Task Model Resolution
vid2vid Cityscapes, segmentation to video download 1024x512
fs-vid2vid FaceForensics, landmarks to video download 512x512

World-to-world Translation

Algorithm Name Task Model Resolution
wc-vid2vid Cityscapes, segmentation to video download 1024x512
GANcraft Semantic block worlds to realistic worlds download arbitrary