Title:Memory-Guided Collaborative Attention for Nighttime Thermal Infrared Image Colorization

Abstract:Nighttime thermal infrared (NTIR) image colorization, also known as translation of NTIR images into daytime color images (NTIR2DC), is a promising research direction to facilitate nighttime scene perception for humans and intelligent systems under unfavorable conditions (e.g., complete darkness). However, previously developed methods have poor colorization performance for small sample classes. Moreover, reducing the high confidence noise in pseudo-labels and addressing the problem of image gradient disappearance during translation are still under-explored, and keeping edges from being distorted during translation is also challenging. To address the aforementioned issues, we propose a novel learning framework called Memory-guided cOllaboRative atteNtion Generative Adversarial Network (MornGAN), which is inspired by the analogical reasoning mechanisms of humans. Specifically, a memory-guided sample selection strategy and adaptive collaborative attention loss are devised to enhance the semantic preservation of small sample categories. In addition, we propose an online semantic distillation module to mine and refine the pseudo-labels of NTIR images. Further, conditional gradient repair loss is introduced for reducing edge distortion during translation. Extensive experiments on the NTIR2DC task show that the proposed MornGAN significantly outperforms other image-to-image translation methods in terms of semantic preservation and edge consistency, which helps improve the object detection accuracy remarkably.