DOLG: Single-Stage Image Retrieval with Deep Orthogonal Fusion of Local and Global Features (ICCV 2021)

Pipeline

Performances

Codes

Requirements

• NVIDIA GPU, Linux, Python3(tested on 3.6.10)
• Tested with CUDA 10.2, cuDNN 7.1 and PyTorch 1.4.0

pip install -r requirements.txt

Training

1. Find datasets via symlinks from datasets/data to the actual locations where the dataset images and annotations are stored. Refer to DATA.md.

2. Set datapath, model, training parameters in configs/resnet101_delg_8gpu.yaml and run job.sh.

Evaluation

1. Feature extraction, set ${total_num} = n * (gpu_cards) in configs/resnet101_delg_8gpu.yaml and run evaler/run.sh for feature extraction.

2. Eval on ROxf and RPar, refer README.md for data fetch and description. Groudtruth file and some examples are prepared in revisitop.

Wights

GLDv2-clean

• R-50-DOLG
• R-101-DOLG

Citation

If the project helps your research, please consider citing our paper as follows.

@inproceedings{yang2021dolg,
  title = {DOLG: Single-Stage Image Retrieval with Deep Orthogonal Fusion of Local and Global Features},
  author = {Min Yang and Dongliang He and Miao Fan and Baorong Shi and Xuetong Xue and Fu Li and Errui Ding and Jizhou Huang},
  booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
  year = {2021}
}

References

pycls(https://github.com/facebookresearch/pycls) pymetric(https://github.com/feymanpriv/pymetric) DELG(https://github.com/feymanpriv/DELG) Parsing-R-CNN(https://github.com/soeaver/Parsing-R-CNN)