DynamicISTD Benchmark
收藏资源简介:
DynamicISTD Benchmark数据集是一个专为无人机红外小目标设计的跨域动态退化数据集,旨在模拟实际应用中遇到的分布偏移。该数据集包含来自不同域的训练和测试集,用于解决跨域场景下红外小目标检测的挑战。数据集的创建过程采用了Wavelet-guided Cross-domain Synthesis策略,通过多频段小波滤波准确分离目标背景,并通过Real-domain Noise Invariance Learning提取目标域的真实噪声特征,构建动态噪声库。DynamicISTD Benchmark数据集的应用领域在于无人机红外小目标检测,旨在解决跨域偏移和异方差噪声扰动的问题。
The DynamicISTD Benchmark is a cross-domain dynamic degradation dataset specifically tailored for UAV-based infrared small target detection, intended to simulate the distribution shifts encountered in real-world applications. This dataset comprises training and test subsets from various domains, designed to tackle the challenges of infrared small target detection in cross-domain scenarios. The dataset is developed using the Wavelet-guided Cross-domain Synthesis strategy: it accurately separates targets from backgrounds through multi-band wavelet filtering, extracts real noise features of the target domain via Real-domain Noise Invariance Learning, and constructs a dynamic noise library. The primary application of the DynamicISTD Benchmark is UAV-based infrared small target detection, where it addresses the issues of cross-domain distribution shift and heteroscedastic noise disturbance.

- 1通过School of Information Engineering, Guangdong University of Technology · 2025年



