Image To Image Translation On Cityscapes

Metrics

FID

Per-pixel Accuracy

mIoU

Results

Performance results of various models on this benchmark

				Paper Title	Repository
CRN	104.7	77.1%	52.4	Photographic Image Synthesis with Cascaded Refinement Networks	-
pix2pixHD	95	81.4%	58.3	High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs
Pix2PixHD-AUG	72.7	-	58	Improving Augmentation and Evaluation Schemes for Semantic Image Synthesis	-
SPADE	71.8	81.9%	62.3	Semantic Image Synthesis with Spatially-Adaptive Normalization
SB-GAN	60.39	-	-	Semantic Bottleneck Scene Generation
SPADE + FFL	59.5	82.5%	64.2	Focal Frequency Loss for Image Reconstruction and Synthesis
CC-FPSE	54.3	82.3%	65.5	Learning to Predict Layout-to-image Conditional Convolutions for Semantic Image Synthesis
SPADE + SESAME	54.2	82.5%	66	SESAME: Semantic Editing of Scenes by Adding, Manipulating or Erasing Objects
USIS	53.67	-	44.78	USIS: Unsupervised Semantic Image Synthesis
CC-FPSE-AUG	52.1	-	63.1	Improving Augmentation and Evaluation Schemes for Semantic Image Synthesis	-
USIS-Wavelet	50.14	-	42.32	Wavelet-based Unsupervised Label-to-Image Translation
SIMS	49.7	75.5%	47.2	Semi-parametric Image Synthesis
OASIS	47.7	-	69.3	You Only Need Adversarial Supervision for Semantic Image Synthesis
DP-GAN	44.1	-	73.6	Dual Pyramid Generative Adversarial Networks for Semantic Image Synthesis
DP-SIMS (ConvNext-L)	38.2	-	76.3	Unlocking Pre-trained Image Backbones for Semantic Image Synthesis	-
INADE	-	-	-	Diverse Semantic Image Synthesis via Probability Distribution Modeling
pix2pix	-	71.0	-	Image-to-Image Translation with Conditional Adversarial Networks
BiGAN	-	19%	-	Adversarially Learned Inference
CoGAN	-	40%	-	Coupled Generative Adversarial Networks
SimGAN	-	20%	-	Learning from Simulated and Unsupervised Images through Adversarial Training

0 of 21 row(s) selected.

Command Palette

Image To Image Translation On Cityscapes

Metrics

Results