A survey on Face Sketch to Photo Translation based on Hybrid Optimization Algorithm and Deep Learning Technique

The objective of sketch-based face image generation is to produce synthetic, highly consistent, and photorealistic face images from input sketches. These images can be used in a variety of fields, including criminal investigation, human-computer interface, entertainment, and virtual human design. The Generative Adversarial Networks (GANs) helps to find the real images from the sketches. It produces a realistic and high-quality face photo with attributes identical to those of the input sketch. The study shows that the different methods are available for the real image generation from the raw data or raw sketch available. The study also shows that the evolutionary method and upgradation in recent techniques are more prominent than the other existing methods. The development of Neural Radiance Fields (NeRFs) and a deep learning approaches, are powerful implicit 3D representation. This study also demonstrates that even the Frequency-consistent (FC) loss is there it can be applied to different generative model for the overall performance, the generated high-fidelity and high-quality images corresponding to the text description and allowed the editing of real images.