Deep pelican based synthesis model for photo-sketch face synthesis and recognition

Face photo-sketch synthesis has a number of uses in digital entertainment in addition to its security applications. In the upcoming 10 years, recent advances in IoT technology are expected to become widespread. Photo sketch synthesis is still a difficult problem to solve due to the distinct qualities of photo and sketch. In this research, a photo to sketch and sketch to photo transformation problem is considered and investigated by the newly well-liked generative models. Traditional GAN-based methods have shown significant improvements, especially in the area of translation problems. However, they are known to be limited in their ability to produce realistic, high-resolution images. Photo-sketch synthesis (PSS) and Photo-sketch recognition (PSR) are two fascinating and related challenges that will be covered in this research. In order to do this, this research offers a novel synthesis framework called the Pelican generative self-attention adversarial synthesis model (PGS-ASM), which creates pictures of varying resolutions in an adversarial manner sequentially from low to high resolution. In addition, the minimization of additional loss functions affords suitable regularization with high-quality and high-resolution PSS. Using two well-known datasets, all findings are analyzed, and the proposed method has significantly improved image quality and PS matching accuracy. The reconstruction error, qualitative measure of structural similarity index (SSIM), feature similarity index (FSIM) and recognition accuracy are examined and compared with traditional adversarial models in the experimental scenario. In addition to this, the qualitative analysis under varying lightning, pose variations and Occlusions are evaluated also the ablation study and cross fold validation are conducted.