The official implementation of the paper.
- Download the dataset from the repository https://github.com/VisDrone/DroneVehicle, then run the following code to crop the white borders:
python tools/data_process.py- Run the following code to process the labels (since the original labels for the "freight-car" category are inconsistent and contain errors such as "*", we have unified them to "freight-car" in the code):
python tools/VOC2DOTA.pypretrained weights: BaiduYun [code: jwqx]
CUDA==11.8
Pytorch==2.1.2
mmcv==2.1.0
mmdet==3.3.0
mmengine==0.10.5
numpy==1.26.4
You can follow the steps below to create an virtual environment:
- install all dependencies:
conda create -n dmm python=3.10
conda activate dmm
conda install pytorch==2.1.2 torchvision==0.16.2 torchaudio==2.1.2 pytorch-cuda=11.8 -c pytorch -c nvidia
pip install -U openmim
mim install mmdet
pip install numpy==1.26.4
- You might encounter the following, Downgrade the pip version to 24.0 (pip install pip==24.0)
Ignoring mmcv: markers 'extra == "mim"' don't match your environment
Ignoring mmengine: markers 'extra == "mim"' don't match your environment
-
Follow the https://github.com/MzeroMiko/VMamba Getting Started Step 2, install selective_scan==0.0.2
-
Clone the code and install:
git clone https://github.com/Another-0/DMM
cd DMM
pip install -v -e .
- train
python ./tools/train.py ${CONFIG_FILE}
- test
python ./tools/test.py ${CONFIG_FILE} ${CHECKPOINT}
For more command-line arguments, please refer to the code details.
Our codes are mainly based on MMRotate and VMamba. Many thanks to the authors!
Please cite our work if you find our work and codes helpful for your research.
@article{zhou2024dmm,
title={DMM: Disparity-guided Multispectral Mamba for Oriented Object Detection in Remote Sensing},
author={Zhou, Minghang and Li, Tianyu and Qiao, Chaofan and Xie, Dongyu and Wang, Guoqing and Ruan, Ningjuan and Mei, Lin and Yang, Yang},
journal={arXiv preprint arXiv:2407.08132},
year={2024}
}