segformer

Introduction

Official Repo

Code Snippet

SegFormer (NeurIPS'2021)

@article{xie2021segformer,
    title={SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers},
    author={Xie, Enze and Wang, Wenhai and Yu, Zhiding and Anandkumar, Anima and Alvarez, Jose M and Luo, Ping},
    journal={arXiv preprint arXiv:2105.15203},
    year={2021}
}

Results

ADE20k

Backbone	Pretrain	Crop Size	Schedule	Train/Eval Set	mIoU	Download
MIT-B0	ImageNet-1k-224x224	512x512	LR/POLICY/BS/EPOCH: 0.00006/poly/16/130	train/val	37.57%	cfg | model | log
MIT-B1	ImageNet-1k-224x224	512x512	LR/POLICY/BS/EPOCH: 0.00006/poly/16/130	train/val	42.25%	cfg | model | log
MIT-B2	ImageNet-1k-224x224	512x512	LR/POLICY/BS/EPOCH: 0.00006/poly/16/130	train/val	46.35%	cfg | model | log
MIT-B3	ImageNet-1k-224x224	512x512	LR/POLICY/BS/EPOCH: 0.00006/poly/16/130	train/val	48.31%	cfg | model | log
MIT-B4	ImageNet-1k-224x224	512x512	LR/POLICY/BS/EPOCH: 0.00006/poly/16/130	train/val	48.59%	cfg | model | log
MIT-B5	ImageNet-1k-224x224	512x512	LR/POLICY/BS/EPOCH: 0.00006/poly/16/130	train/val	49.61%	cfg | model | log

More

You can also download the model weights from following sources:

• BaiduNetdisk: https://pan.baidu.com/s/1gD-NJJWOtaHCtB0qHE79rA with access code s757