High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs

CVPR 2018 Ting-Chun WangMing-Yu LiuJun-Yan ZhuAndrew TaoJan KautzBryan Catanzaro

We present a new method for synthesizing high-resolution photo-realistic images from semantic label maps using conditional generative adversarial networks (conditional GANs). Conditional GANs have enabled a variety of applications, but the results are often limited to low-resolution and still far from realistic... (read more)

PDF Abstract

Evaluation results from the paper


Task Dataset Model Metric name Metric value Global rank Compare
Image-to-Image Translation ADE20K Labels-to-Photos pix2pixHD mIoU 20.3 # 3
Image-to-Image Translation ADE20K Labels-to-Photos pix2pixHD Accuracy 69.2% # 2
Image-to-Image Translation ADE20K Labels-to-Photos pix2pixHD FID 81.8 # 3
Image-to-Image Translation ADE20K-Outdoor Labels-to-Photos pix2pixHD mIoU 17.4 # 2
Image-to-Image Translation ADE20K-Outdoor Labels-to-Photos pix2pixHD Accuracy 71.6% # 3
Image-to-Image Translation ADE20K-Outdoor Labels-to-Photos pix2pixHD FID 97.8 # 3
Image-to-Image Translation Cityscapes Labels-to-Photo pix2pixHD Class IOU # 6
Image-to-Image Translation Cityscapes Labels-to-Photo pix2pixHD Per-class Accuracy # 5
Image-to-Image Translation Cityscapes Labels-to-Photo pix2pixHD Per-pixel Accuracy 81.4% # 2
Image-to-Image Translation Cityscapes Labels-to-Photo pix2pixHD mIoU 58.3 # 2
Image-to-Image Translation Cityscapes Labels-to-Photo pix2pixHD FID 95 # 3
Image-to-Image Translation COCO-Stuff Labels-to-Photos pix2pixHD mIoU 14.6 # 3
Image-to-Image Translation COCO-Stuff Labels-to-Photos pix2pixHD Accuracy 45.8% # 2
Image-to-Image Translation COCO-Stuff Labels-to-Photos pix2pixHD FID 111.5 # 3