Zhenliang Ni, Xinghao Chen, Yingjie Zhai, Yehui Tang, and Yunhe Wang
- 2024/07/01: The paper of CGRSeg is accepted by ECCV 2024.
- 2024/05/10: Codes of CGRSeg are released in Pytorch and paper in [arXiv].
The overall architecture of CGRSeg. The Rectangular Self-Calibration Module (RCM) is designed for spatial feature reconstruction and pyramid context extraction. The rectangular self-calibration attention (RCA) explicitly models the rectangular region and calibrates the attention shape. The Dynamic Prototype Guided (DPG) head is proposed to improve the classification of the foreground objects via explicit class embedding.
COCO-Stuff-10k
Pascal Context
-
conda create --name ssa python=3.8 -y conda activate ssa pip install torch==1.8.2+cu102 torchvision==0.9.2+cu102 torchaudio==0.8.2 pip install timm==0.6.13 pip install mmcv-full==1.6.1 pip install opencv-python==4.1.2.30 pip install "mmsegmentation==0.27.0"CGRSeg is built based on mmsegmentation-0.27.0, which can be referenced for data preparation.
-
Train
# Single-gpu training python train.py local_configs/cgrseg/cgrseg-t_ade20k_160k.py # Multi-gpu (4-gpu) training bash dist_train.sh local_configs/cgrseg/cgrseg-t_ade20k_160k.py 4
-
Test
# Single-gpu testing python test.py local_configs/cgrseg/cgrseg-t_ade20k_160k.py ${CHECKPOINT_FILE} --eval mIoU # Multi-gpu (4-gpu) testing bash dist_test.sh local_configs/cgrseg/cgrseg-t_ade20k_160k.py ${CHECKPOINT_FILE} 4 --eval mIoU
If you find CGRSeg useful in your research or applications, please consider giving a star ⭐ and citing using the following BibTeX:
@article{ni2024context,
title={Context-Guided Spatial Feature Reconstruction for Efficient Semantic Segmentation},
author={Ni, Zhenliang and Chen, Xinghao and Zhai, Yingjie and Tang, Yehui and Wang, Yunhe},
journal={arXiv preprint arXiv:2405.06228},
year={2024}
}