Awesome

Tiny Object Detection in Remote Sensing Images Based on Object Reconstruction and Multiple Receptive Field Adaptive Feature Enhancement (IEEE TGRS 2024)

This repository contains python implementation of our paper ORFENet.

1. Required environments:

Linux
Python 3.6+
PyTorch 1.3+
CUDA 9.2+
GCC 5+
MMCV
cocoapi-aitod

2. Install and start ORFENet:

Note that our ORFENet is based on the MMDetection 2.24.1. Assume that your environment has satisfied the above requirements, please follow the following steps for installation.

git clone https://github.com/dyl96/ORFENet.git
cd ORFENet
pip install -r requirements/build.txt
python setup.py develop

Prepare Dataset:

Download AI-TODv2 dataset; Download LEVIR-Ship dataset.

Train and test：

Train aitodv2 dataset:

python tools/train.py configs_orfenet/orfenet/aitodv2_fcos_r50_p2_hrfe_or_3x.py

Train LEVIR-Ship dataset:

python tools/train.py configs_orfenet/orfenet/levir_ship_fcos_r50_p2_hrfe_or_1x.py

Test LEVIR-Ship dataset:

python tools/test.py configs_orfenet/orfenet/levir_ship_fcos_r50_p2_hrfe_or_1x.py work_dirs/levir_ship_fcos_r50_p2_hrfe_or_1x/epoch_12.pth --eval bbox

Checkpoint Download：

Baidu Pan：https://pan.baidu.com/s/1eyJiSV12hX6gggiuq8-DFA?pwd=uon2 code：uon2

3. Visual Results

<img src="assets/1.png"/> Visual comparisons of the proposed method and other methods. (a) the baseline FCOS. (b) FSANet. (c) Cascade-R-CNN w/ NWD-RKA. (d) The proposed ORFENet. The green boxes denote the true positive predictions, the red boxes denote the false negative predictions, and the blue boxes denote the false positive predictions.

4. Citation

Please cite our paper if you find the work useful:

@ARTICLE{10462223,
  author={Liu, Dongyang and Zhang, Junping and Qi, Yunxiao and Wu, Yinhu and Zhang, Ye},
  journal={IEEE Geoscience and Remote Sensing Letters}, 
  title={A Tiny Object Detection Method Based on Explicit Semantic Guidance for Remote Sensing Images}, 
  year={2024},
  volume={},
  number={},
  pages={1-1},
  doi={10.1109/LGRS.2024.3374418}}