The implementation of [DASR]

Version info: pytorch 2.0.0, python 3.9

Steps

Modify： Define your path in configs/dasr.yaml

Choose: Choose parameter in configs/dasr.yaml

parameter	explanation
detect_mode	Run single pic or dataset
feature_save	Whether to save feature data
plt_mode	Whether to show / save result by plt
display_mode	Display result by bounding box or response map
dataset	Run which dataset
transform_mode	Normalize by different mean and std
model_select	Process query or reference
use_nms	Whether to use nms
use_DASR*	Whether to use dasr*
use_rank	Whether to use rank
backbone	Choose pre-train parameter for backbone
feature_map_pooling	Whether to use pooling on feature map

Run:

python main.py 

Reference

@article{xiao2022deeply,
  title={Deeply activated salient region for instance search},
  author={Xiao, Hui-Chu and Zhao, Wan-Lei and Lin, Jie and Hong, Yi-Geng and Ngo, Chong-Wah},
  journal={ACM Transactions on Multimedia Computing, Communications and Applications},
  volume={18},
  number={3s},
  pages={1--19},
  year={2022},
  publisher={ACM New York, NY}
}

Publisher

Yi-Bo Miao Xiamen University