High Performance Cityscape Semantic Segmentaion

We will also release the all the state-of-the-art models and code trained on Cityscape dataset including Deeplabv3, Deeplabv3+, PSPnet, DAnet.

GALD-Net (BMVC 2019)

We propose Global Aggregation then Local Distribution (GALD) scheme to distribute global information to each position adaptively according to the local information around the position. GALD net achieves top performance on Cityscapes dataset. Both source code and models will be available soon. The work was done at DeepMotion AI Research

GFF-Net (arxiv)

We proposed Gated Fully Fusion (GFF) to fuse features from multiple levels through gates in a fully connected way. Specifically, features at each level are enhanced by higher-level features with stronger semantics and lower-level features with more details, and gates are used to control the pass of useful information which significantly reducing noise propagation during fusion. (Joint work: Key Laboratory of Machine Perception, School of EECS @Peking University and DeepMotion AI Research )

DGCNet (BMVC 2019)

We propose Dual Graph Convolutional Network (DGCNet) models the global context of the input feature by modelling two orthogonal graphs in a single framework. (Joint work: University of Oxford, Peking University and DeepMotion AI Research)

Comparisons with state-of-the-art models on Cityscapes dataset

Method	Conference	Backbone	mIoU(%)
RefineNet	CVPR2017	ResNet-101	73.6
SAC	ICCV2017	ResNet-101	78.1
PSPNet	CVPR2017	ResNet-101	78.4
DUC-HDC	WACV2018	ResNet-101	77.6
AAF	ECCV2018	ResNet-101	77.1
BiSeNet	ECCV2018	ResNet-101	78.9
PSANet	ECCV2018	ResNet-101	80.1
DFN	CVPR2018	ResNet-101	79.3
DSSPN	CVPR2018	ResNet-101	77.8
DenseASPP	CVPR2018	DenseNet-161	80.6
OCNet	-	ResNet-101	81.7
CCNet	ICCV2019	ResNet-101	81.4
GALD-Net	BMVC2019	ResNet50	80.8
GALD-Net	BMVC2019	ResNet101	81.8
GFF-Net	-	ResNet101	82.3
DGCN-Net	BMVC2019	ResNet101	82.0
GALD-Net(use coarse data)	BMVC2019	ResNet101	82.9
GALD-Net(use Mapillary)	BMVC2019	ResNet101	83.3

Detailed Results are shown

GALD-Net: here
GFF-Net:here
Both are (Single Model Result)

Citation

Please read our paper for model details

@inproceedings{xiangtl_gald
title={Global Aggregation then Local Distribution in Fully Convolutional Networks},
author={Li, Xiangtai and Zhang, Li and You, Ansheng and Yang, Maoke and Yang, Kuiyuan and Tong, Yunhai},
booktitle={British Machine Vision Conference},
year={2019}
}

@inproceedings{xiangtl_gff
  title     = {GFF: Gated Fully Fusion for semantic segmentation},
  author    = {Xiangtai Li and Houlong Zhao and Lei Han and Yunhai Tong and Kuiyuan Yang},
  booktitle = {arXiv preprint arXiv:1903.11816},
  year = {2019}
}

@inproceedings{zhangli_dgcn
title={Dual Graph Convolutional Network for Semantic Segmentation},
author={Zhang, Li(*) and Li, Xiangtai(*) and Arnab, Anurag and Yang, Kuiyuan and Tong, Yunhai and Torr, Philip HS},
booktitle={British Machine Vision Conference},
year={2019}
}

License

MIT License