TVA-GAN: attention guided generative adversarial network for thermal to visible image transformations

In the recent improvement in deep learning approaches for realistic image generation and translation, Generative Adversarial Networks (GANs) delivered favorable results. GAN generates novel samples that look indistinguishable from authentic images. This paper proposes a generative network thermal-to-visible translation. Thermal to Visible synthesis is challenging due non-availability of accurate semantic textural information thermal The sensors acquire face images by capturing object’s luminance with fewer details about actual facial information. However, it advantageous low-light night-time vision, where cannot be captured complex environment an RGB camera. We design new Attention-guided Cyclic Network Face transformation (TVA-GAN) integrating attention network. utilize guidance recurrent block Inception module simplify space toward optimum solution. proposed TVA-GAN trained evaluated visible over three benchmark datasets, including WHU-IIP, Tufts Thermal2RGB, CVBL-CHILD datasets. results show promising compared state-of-the-art methods. For TVA-GAN, code available at: https://github.com/GANGREEK/TVA-GAN .