Pinwheel-shaped Convolution and Scale-based Dynamic Loss for Infrared Small Target Detection

Authors: Jiangnan Yang, Shuangli Liu, Jingjun Wu, Xinyu Su, Nan Hai, Xueli Huang

AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We construct a new benchmark, SIRST-UAVB, which is the largest and most challenging dataset to date for realshot single-frame infrared small target detection. Lastly, by integrating PConv and SD Loss into the latest small target detection algorithms, we achieved significant performance improvements on IRSTD-1K and our SIRST-UAVB dataset, validating the effectiveness and generalizability of our approach.
Researcher Affiliation Academia 1School of Information and Engineering, Southwest University of Science and Technology 2School of Electronic and Optical Engineering, Nanjing University of Science and Technology EMAIL
Pseudocode No The paper describes methods using mathematical equations and figures for architecture visualization (e.g., Figure 3: Architecture of the pinwheel-shaped convolutional module), but does not contain explicit pseudocode or algorithm blocks.
Open Source Code Yes Code https://github.com/JN-Yang/PConv-SDloss-Data
Open Datasets Yes Existing IRSTDS datasets (Dai et al. 2021a; Zhang et al. 2022) have a low proportion of small targets... To address these limitations, we developed a new single-frame IRSTDS dataset, SIRST-UAVB, capturing unmanned aerial vehicle (UAV) and birds. Our contributions can be summarized as follows: We construct SIRST-UAVB, the largest publicly available dataset for real IRSTDS, encompassing comprehensive spatial domain challenges. Datasets. To evaluate the impact of PConv and SD loss across different hyperparameters on targets of varying scales, we used two datasets: IRSTD-1K (Zhang et al. 2022), containing 1,000 real infrared images with an average larger target scale and a resolution of 512 × 512 pixels, and our SIRST-UAVB, which features smaller targets.
Dataset Splits Yes Both datasets were split into training and testing sets with a 4:1 ratio.
Hardware Specification Yes We conducted ablation experiments on IRST detection and segmentation models, using the PyTorch framework on RTX3090 GPUs.
Software Dependencies No The paper mentions using the 'PyTorch framework' but does not provide a specific version number. No other software dependencies with version numbers are listed.
Experiment Setup Yes Implementation Detail. We conducted ablation experiments on IRST detection and segmentation models, using the PyTorch framework on RTX3090 GPUs. For detection models, the input image size was set to 640, batch size to 64, epochs to 700, patience to 70, and learning rate to 0.01. For segmentation models, the input image size was set to 256, batch size to 4, epochs to 400, and learning rate to 0.05.