AugTarget Data Augmentation

Datasets

MDFA dataset is available at MDvsFa cGan.
The SIRST Augment dataset: download from Google Drive or BaiduYun Drive with code ojd4.

Usage

Train

python train.py --net-name agpcnet_1 --batch-size 8 --save-iter-step 20 --dataset mdfa --learning-rate 0.05

python train.py --net-name agpcnet_1 --batch-size 8 --save-iter-step 40 --dataset sirstaug --learning-rate 0.1

python train.py --net-name agpcnet_1 --batch-size 8 --save-iter-step 100 --dataset merged --learning-rate 0.05

Inference

python inference.py --pkl-path {checkpoint path} --image-path {image path}

Evaluation

python evaluation.py --dataset {dataset name} 
                     --sirstaug-dir {base dir of sirstaug}
                     --mdfa-dir {base dir of MDFA}
                     --pkl-path {checkpoint path}

Results

Methods	Data	Precision	Recall	mIoU	Fmeasure	AUC	Download
AGPCNet	MDFA	0.5939	0.7241	0.4843	0.6525	0.8682	model
AGPCNet	SIRST Aug	0.8323	0.8542	0.7288	0.8431	0.9344	model
AGPCNet	Merged	0.7453	0.8384	0.6517	0.7891	0.9194	model
AGPCNet+AugTarget	MDFA	0.6482	0.7141	0.5146	0.6795	0.8699	model
AGPCNet+AugTarget	SIRST Aug	0.8449	0.8704	0.7505	0.8574	0.9378	model
AGPCNet+AugTarget	Merged	0.7576	0.8658	0.6780	0.8081	0.9395	model

Acknowledgement

Our Target Augmentation algorithm is based on the random strategy of Random-Erasing, thanks for their contributions. This repository is based on baseline model from AGPCNet and modified part of the code.

Contact

If any questions, kindly contact with Shengjia Chen via e-mail: csj_uestc@126.com.

References

Zhong, Zhun, et al. "Random erasing data augmentation." Proceedings of the AAAI conference on artificial intelligence. Vol. 34. No. 07. 2020. [code]
Zhang, Tianfang, et al. "AGPCNet: Attention-Guided Pyramid Context Networks for Infrared Small Target Detection." arXiv preprint arXiv:2111.03580 (2021). [code]

Citation

If you find this repo useful, please cite our paper.

@inproceedings{chen2023augtarget,
  title={AugTarget Data Augmentation for Infrared Small Target Detection},
  author={Chen, Shengjia and Zhu, Jiewen and Ji, Luping and Pan, Hongjun and Xu, Yuhao},
  booktitle={IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  pages={1--5},
  year={2023},
  organization={IEEE}
}

UESTC-nnLab/AugTarget