In this paper, we present a novel method for image super-resolution (SR) employing a Dual-Stage Generative Adversarial Network (DS-GAN). After training using a set of low-resolution (LR) and high-resolution (HR) images, we obtain the SR for the given test LR image. Our approach involves training two consecutive GANs to enhance high-resolution fea¬tures by constraining the discriminators to match the high as well as the reconstructed low-resolution images. To take care of global as well as local characteristics of generated SR, the generator loss for both GANs are modified to include the already available perceptual loss as well as a new loss formed using the Inhomogeneous Gaussian Markov Random Field (IGMRF) prior. The addition of IGMRF prior loss helps us in regularing the solution better. While the perceptual loss can preserve the global SR features, the local attributes, i.e., smoothness and discontinuities among SR pixels in generated images are taken care by the IGMRF prior. The advantage of training a dual architecture is that once trained for a magnification factor of 2, we can use the same parameters to get higher magnifications. With this, the proposed approach has performed better even for a magnification factor of 8, thus reducing the computational burden. We evaluate results on the DIVerse 2K (DIV2K) and benchmark Set14 datasets for 4× and 8× SR using qualitative as well as quantitative measures. Comparison with those proposed using the SRResNet architecture, the ZZPM team, DRCT, and including diffusion-based approaches demonstrate the efficacy of the proposed approach.

Electric and hybrid electric vehicles (EV/HEV) are promising solutions for fossil fuel conservation and pollution reduction for a safe environment and sustainable transportation. The design of these energy-efficient powertrains requires optimization of components, systems, and controls. Controls entail battery management, fuel consumption, driver performance demand emissions, and management strategy. The hardware optimization entails powertrain architecture, transmission type, power electronic converters, and energy storage systems. In this overview, all these factors are addressed and reviewed. Major challenges and future technologies for EV/HEV are also discussed. Published suggestions and recommendations are surveyed and evaluated in this review. The outcomes of detailed studies are presented in tabular form to compare the strengths and weaknesses of various methods. Furthermore, issues in the current research are discussed, and suggestions toward further advancement of the technology are offered. This article analyzes current research and suggests challenges and scope of future research in EV/HEV and can serve as a reference for those working in this field.