Revisiting the Scale Loss Function and Gaussian-Shape Convolution for Infrared Small Target Detection

TL;DR

This paper improves infrared small target detection by proposing a stable scale loss function and a Gaussian-shaped convolution that better captures target characteristics, leading to enhanced detection performance.

Contribution

It introduces a diff-based scale loss for stable training and a Gaussian-shaped convolution with adaptive orientation for better spatial attention in infrared small target detection.

Findings

Achieved consistent improvements in mIoU, P_d, and F_a on multiple datasets.

Demonstrated the effectiveness of the proposed loss and convolution in extensive experiments.

Analyzed geometric properties of scale loss variants to understand detection behavior.

Abstract

Infrared small target detection still faces two persistent challenges: training instability from non-monotonic scale loss functions, and inadequate spatial attention due to generic convolution kernels that ignore the physical imaging characteristics of small targets. In this paper, we revisit both aspects. For the loss side, we propose a \emph{diff-based scale loss} that weights predictions according to the signed area difference between the predicted mask and the ground truth, yielding strictly monotonic gradients and stable convergence. We further analyze a family of four scale loss variants to understand how their geometric properties affect detection behavior. For the spatial side, we introduce \emph{Gaussian-shaped convolution} with a learnable scale parameter to match the center-concentrated intensity profile of infrared small targets, and augment it with a \emph{rotated pinwheel mask} that adaptively aligns the kernel with target orientation via a straight-through estimator. Extensive experiments on IRSTD-1k, NUDT-SIRST, and SIRST-UAVB demonstrate consistent improvements in $mIoU$, $P_d$, and $F_a$ over state-of-the-art methods. We release our anonymous code and pretrained models.