Home

Awesome

Background Semantics Matter: Cross-Task Feature Exchange Network for Clustered Infrared Small Target Detection With Sky-Annotated Dataset

Paper link: Background Semantics Matter: Cross-Task Feature Exchange Network for Clustered Infrared Small Target Detection With Sky-Annotated Dataset

The repository of BAFE-Net: https://github.com/GrokCV/BAFE-Net

Introduction

You can download our DenseSIRST dataset from Google Drive.<br>

Detecting small, low-contrast objects against complex backgrounds in infrared images is a major challenge. Existing datasets mainly feature sparsely distributed objects and lack pixel-wise background annotations. This is in stark contrast to the dense, tiny objects often encountered in the real world. To bridge this gap, we propose DenseSIRST: a carefully annotated infrared image dataset filled with densely clustered small objects, a more realistic and challenging benchmark for algorithm training and evaluation.We shows a set of images from the DenseSIRST dataset, demonstrating the dense distribution of small objects and the corresponding sky segmentation annotations. These annotations provide valuable background information for developing and evaluating detection algorithms.

DatasetsImage TypeAnnotation TypeImage NumberTarget NumberAverage Target AreaSparse or ClusteredBackground Semantic Annotation
SIRST V1RealPixel42753323Sparse×
SIRST V2RealPixel + BBox + Point102464824Sparse×
IRSTD1KRealPixel1001149538Sparse×
SIRSTAUGSyntheticPixel8525927888Sparse×
DenseSIRST (Ours)Real + SyntheticPixel + BBox + Point1024136556Cluster

DenseSIRST

Dataset Composition

Annotations and Labels

Statistical Characteristics

<div align="center"> <img src="./docs/object_sizes.png" width="49.5%"> </div> <p float="left"> <img src="./docs/image_lcm.png" width="49.5%" /> <img src="./docs/image_brightness.png" width="49.5%" /> </p>

Dataset Structure

For both training and inference, the following dataset structure is required:

|- data
    |- SIRSTdevkit
        |-PNGImages
            |-Misc_1.png
            ......
        |-SIRST
            |-BBox
                |-Misc_1.xml
                ......
            |-BinaryMask
                |-Misc_1_pixels0.png
                |-Misc_1.xml
                ......
            |-PaletteMask
                |-Misc_1.png
                ......
            |-Point_label
                |-Misc_1_pixels0.txt
                ......
        |-SkySeg
            |-BinaryMask
                |-Misc_1_pixels0.png
                |-Misc_1.xml
                ......
            |-PaletteMask
                |-Misc_1.png
                ......
        |-Splits
            |-train_v2.txt
            |-test_v2.txt
            ......

Please make sure that the path of your data set is consistent with the data_root in configs/detection/_base_/datasets/sirst_det_seg_voc_skycp.py

Citation

If you use our dataset or code in your research, please cite this project.

@article{xiao2024bafenet,
	title={Background Semantics Matter: Cross-Task Feature Exchange Network for Clustered Infrared Small Target Detection With Sky-Annotated Dataset}, 
	author={Mengxuan Xiao and Qun Dai and Yiming Zhu and Kehua Guo and Huan Wang and Xiangbo Shu and Jian Yang and Yimian Dai},
	year={2024},
	journal={arXiv preprint arXiv:2407.20078},
}

@article{dai2023one,
  title={One-stage cascade refinement networks for infrared small target detection},
  author={Dai, Yimian and Li, Xiang and Zhou, Fei and Qian, Yulei and Chen, Yaohong and Yang, Jian},
  journal={IEEE Transactions on Geoscience and Remote Sensing},
  volume={61},
  pages={1--17},
  year={2023},
}