SpirDet: Towards Efficient, Accurate and Lightweight Infrared Small Target Detector

Qianchen Mao,Qiang Li,Bingshu Wang,Yongjun Zhang,Tao Dai,C.L. Philip Chen

2024-02-08

Abstract:In recent years, the detection of infrared small targets using deep learning methods has garnered substantial attention due to notable advancements. To improve the detection capability of small targets, these methods commonly maintain a pathway that preserves high-resolution features of sparse and tiny targets. However, it can result in redundant and expensive computations. To tackle this challenge, we propose SpirDet, a novel approach for efficient detection of infrared small targets. Specifically, to cope with the computational redundancy issue, we employ a new dual-branch sparse decoder to restore the feature map. Firstly, the fast branch directly predicts a sparse map indicating potential small target locations (occupying only 0.5\% area of the map). Secondly, the slow branch conducts fine-grained adjustments at the positions indicated by the sparse map. Additionally, we design an lightweight DO-RepEncoder based on reparameterization with the Downsampling Orthogonality, which can effectively reduce memory consumption and inference latency. Extensive experiments show that the proposed SpirDet significantly outperforms state-of-the-art models while achieving faster inference speed and fewer parameters. For example, on the IRSTD-1K dataset, SpirDet improves $MIoU$ by 4.7 and has a $7\times$ $FPS$ acceleration compared to the previous state-of-the-art model. The code will be open to the public.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

This paper mainly focuses on the problem of infrared small target detection, especially how to efficiently, accurately, and lightly perform this task in deep learning methods. Current methods usually retain high-resolution feature maps to enhance the detection ability of small targets, but this leads to computational redundancy. Therefore, the paper proposes a new method called SpirDet, which uses a dual-branch sparse decoder to reduce computational costs. The fast branch directly predicts the rough position of small targets on low-resolution feature maps, while the slow branch refines the targets on high-resolution maps at these positions. In addition, they design a reparameterization encoder based on downsampling orthogonality (DO-RepEncoder), which effectively reduces memory consumption and inference latency.

Experimental results show that SpirDet achieves faster inference speed and fewer parameters while maintaining high performance. For example, on the IRSTD-1K dataset, SpirDet improves the MIoU by 4.7 and increases the frame rate by 7 times. The paper also compares with the existing best models, demonstrating the superiority of SpirDet in terms of metrics and speed.

In summary, this paper aims to address the efficiency and accuracy issues in infrared small target detection in deep learning. By leveraging innovative network architectures and module designs, it improves detection performance and reduces computational requirements.