SAROD: Efficient End-to-end Object Detection on SAR Images withReinforcement Learning

Yolov3: https://github.com/eriklindernoren/PyTorch-YOLOv3

Yolov5: https://github.com/ultralytics/yolov5

mmdetection: https://github.com/open-mmlab/mmdetection

EfficientObjectDetection: https://github.com/uzkent/EfficientObjectDetection

Overview of our framework.

Clone

git clone https://github.com/anonymous-hub/SAROD

Dataset

HRSID Dataset can be downloaded in here

Pre-Processed dataset for the result can be downloaded by running the file or here.

A example script for downloading the testset is as follows:

# Download the dataset
cd dataset
bash download_HRSID_cropped.sh
cd ..

Download pre-trained model weights

The pretrained weights can be downloaded by running the files or here.

# Download the pre-trained SAROD weights
cd weights
bash download_SAROD_RL_weight.sh
bash download_yolov5_480_weight.sh
bash download_yolov5_96_weight.sh
cd ..

# Download the pre-trained baseline weights
cd weights
bash download_yolov3_480_weight.sh
bash download_yolov3_96_weight.sh
bash download_retinanet_weight.sh
bash download_faster_rcnn_weight.sh
cd ..

Setup

pip install -r requirements.txt

Train

Yolo v3, Yolo v5, Faster-RCNN

refer to 'demo_training.ipynb'

!python train.py --epochs 10\
--detector_batch_size 4\
--device 2\
--test_epoch 1\
--eval_epoch 1\
--step_batch_size 2\
--save_path save

Evaluation

Yolo v3, Yolo v5, Faster-RCNN

refer to 'demo_evaluation.ipynb'

!python test.py --epochs 1\
--detector_batch_size 1\
--device 0\
--test_epoch 1\
--eval_epoch 1\
--rl_weight SAROD_RL\
--h_detector_weight yolov5_480.pt\
--l_detector_weight yolov5_96.p