segmentation

Applying PoolFormer to Semantic Segmentation

Our semantic segmentation implementation is based on MMSegmentation v0.19.0 and PVT segmentation. Thank the authors for their wonderful works.

For details see MetaFormer is Actually What You Need for Vision.

Note

Please note that we just simply follow the hyper-parameters of PVT which may not be the optimal ones for PoolFormer. Feel free to tune the hyper-parameters to get better performance.

Bibtex

@article{yu2021metaformer,
  title={MetaFormer is Actually What You Need for Vision},
  author={Yu, Weihao and Luo, Mi and Zhou, Pan and Si, Chenyang and Zhou, Yichen and Wang, Xinchao and Feng, Jiashi and Yan, Shuicheng},
  journal={arXiv preprint arXiv:2111.11418},
  year={2021}
}

Usage

Install MMSegmentation v0.19.0. Dockerfile_mmdetseg is the docker file that I use to set up the environment for detection and segmentation. You can also refer to it.

Data preparation

Prepare ADE20K according to the guidelines in MMSegmentation.

Results and models

Method	Backbone	Pretrain	Iters	mIoU	Config	Download
Semantic FPN	PoolFormer-S12	ImageNet-1K	40K	37.2	config	log & model
Semantic FPN	PoolFormer-S24	ImageNet-1K	40K	40.3	config	log & model
Semantic FPN	PoolFormer-S36	ImageNet-1K	40K	42.0	config	log & model
Semantic FPN	PoolFormer-M36	ImageNet-1K	40K	42.4	config	log & model
Semantic FPN	PoolFormer-M48	ImageNet-1K	40K	42.7	config	log & model

All the models can also be downloaded by BaiDu Yun (password: esac).

Evaluation

To evaluate PoolFormer-S12 + Semantic FPN on a single node with 8 GPUs run:

dist_test.sh configs/sem_fpn/PoolFormer/fpn_poolformer_s12_ade20k_40k.py /path/to/checkpoint_file 8 --out results.pkl --eval mIoU

Training

To train PoolFormer-S12 + Semantic FPN on a single node with 8 GPUs run:

dist_train.sh configs/sem_fpn/PoolFormer/fpn_poolformer_s12_ade20k_40k.py 8