SMPISD-MTPNet: Scene Semantic Prior-Assisted Infrared Ship Detection Using Multi-Task Perception Networks

[IEEE TGRS] Implementation of our paper "SMPISD-MTPNet: Scene Semantic Prior-Assisted Infrared Ship Detection Using Multi-Task Perception Networks". paper

Requirement

Please check requirements.txt

Datasets

Our dataset is here: https://pan.baidu.com/s/1FwSVOrNgu1XJO6EvucWNGA?pwd=hchc The code is hchc

Pretraining Weights

Our weights is here: https://pan.baidu.com/s/1tXpmKDyMTyMysqC3TaP8_w?pwd=ymsb

The code is ymsb

Commands for Training

• Install the environment according to requirements.txt
• Check the input.xlsx where there are some settings about our training.
• Run launch.py

python launch.py

• Checkpoints and Logs will be saved to training_save

Citation

@ARTICLE{10802996,
  author={Hu, Chen and Dong, Xiaogang and Huang, Yian and Wang, Lele and Xu, Liang and Pu, Tian and Peng, Zhenming},
  journal={IEEE Transactions on Geoscience and Remote Sensing}, 
  title={SMPISD-MTPNet: Scene Semantic Prior-Assisted Infrared Ship Detection Using Multitask Perception Networks}, 
  year={2025},
  volume={63},
  number={},
  pages={1-14},
  keywords={Marine vehicles;Semantics;Feature extraction;Accuracy;Image segmentation;Object detection;Multitasking;Shape;Synthetic aperture radar;Oceans;Gradient-based module;infrared ship detection (IRSD);multitask perception;scene segmentation;scene semantic prior},
  doi={10.1109/TGRS.2024.3516879}}