mobilevit

Introduction

Official Repo

Code Snippet

MobileViT (ICLR'2022)

@article{mehta2021mobilevit,
    title={Mobilevit: light-weight, general-purpose, and mobile-friendly vision transformer},
    author={Mehta, Sachin and Rastegari, Mohammad},
    journal={arXiv preprint arXiv:2110.02178},
    year={2021}
}

MobileViTV2 (ArXiv'2022)

@article{mehta2022separable,
    title={Separable self-attention for mobile vision transformers},
    author={Mehta, Sachin and Rastegari, Mohammad},
    journal={arXiv preprint arXiv:2206.02680},
    year={2022}
}

Results

PASCAL VOC + COCO VOC Subset

Segmentor	Pretrain	Backbone	Crop Size	Schedule	Train/Eval Set	mIoU (EMA)	Download
DeepLabV3	ImageNet-1k-224x224	MobileViT-Small	512x512	LR/POLICY/BS/EPOCH: 0.0009/cosine/64/50	voc trainaug + cocovocsubet train / voc val		cfg | model | log
DeepLabV3	ImageNet-1k-224x224	MobileViTV2-0.50	512x512	LR/POLICY/BS/EPOCH: 0.0005/cosine/64/50	voc trainaug + cocovocsubet train / voc val		cfg | model | log
DeepLabV3	ImageNet-1k-224x224	MobileViTV2-1.00	512x512	LR/POLICY/BS/EPOCH: 0.0005/cosine/64/50	voc trainaug + cocovocsubet train / voc val		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