Borrow from Source Models: Efficient Infrared Object Detection with Limited Examples

Abstract

Recent deep models trained on large-scale RGB datasets lead to considerable achievements in visual detection tasks. However, the training examples are often limited for an infrared detection task, which may deteriorate the performance of deep detectors. In this paper, we propose a transfer approach, Source Model Guidance (SMG), where we leverage a high-capacity RGB detection model as the guidance to supervise the training process of an infrared detection network. In SMG, the foreground soft label generated from the RGB model is introduced as source knowledge to provide guidance for cross-domain transfer. Additionally, we design a Background Suppression Module in the infrared network to receive the knowledge and enhance the foreground features. SMG is easily plugged into any modern detection framework, and we show two explicit instantiations of it, SMG-C and SMG-Y, based on CenterNet and YOLOv3, respectively. Extensive experiments on different benchmarks show that both SMG-C and SMG-Y achieve remarkable performance even if the training set is scarce. Compared to advanced detectors on public FLIR, SMG-Y with 77.0% mAP outperforms others in accuracy, and SMG-C achieves real-time detection at a speed of 107 FPS. More importantly, SMG-Y trained on a quarter of the thermal dataset obtains 74.5% mAP, surpassing most state-of-the-art detectors with full FLIR as training data.

Infrared multi‐target detection and tracking in dense urban traffic scenes

“…Moreover, to enhance recognition accuracy, various methods integrating visible light have been explored. On one front, several infrared object detection algorithms have improved performance through transfer learning [9][10][11][12]. Concurrently, multi-spectral detection algorithms combining visible and infrared imagery [13][14][15][16] have demonstrated noteworthy results on datasets like KAIST [17].…”

“…In this equation, 𝛾 is the aspect ratio of space, 𝜎 is the standard deviation of the Gaussian envelope function, 𝜆 is the wavelength of the cosine function, 𝜙 is the phase offset, and x ′ and y ′ are, respectively, the horizontal and vertical coordinates of the image at point (x, y), as illustrated in Equations ( 8) and (9), where 𝜃 denotes the normal direction of the parallel stripes of the Gabor filter.…”

Computational Intelligence in Remote Sensing

“…The following three papers mainly utilize deep learning techniques to solve practical problems in the field of remote sensing image object detection. The first paper, by R. Chen and S. Liu et al, proposes an effective infrared object detection method based on source model guidance [4]. They show two explicit examples based on Cen-terNet and YOLOv3, respectively, and experimentally demonstrate that the method can achieve powerful performance with limited samples.…”