Analysis of Methods for Recognizing Facial Images From Face Sketch
Abstract
The paper provides an overview of face recognition methodologies using Face sketch in the light of modern contexts and their respective application domains. Face sketch normally find their application in criminology and form a vital part of identification of criminals. Although Face sketch-based recognitions were affected through traditional methods in the recent years, such systems are being upgraded through automation with the aid of artificial intelligence. The article throws light upon sample databases used for Face sketch and their characteristics. The results of the current study have pointed out the differences between conventional methods of recognition and those with the use of deep learning. A comparative analysis of handcrafted features versus deep learning-based features in Face sketch recognition is presented here. Component-based and holistic approaches are compared in this context. Deep learning methods, while giving high accuracy, require a large number of samples and have often given lower results under real-world conditions. Besides, it deals with the importance and variance of GAN for the generation of facial images from Face sketch. GANs allow the generation of high-quality synthetic data, thereby allowing an extension of facial image datasets for Face sketch databases.