The paper concentrates on the latest developments in the field of deep-learning-based image deblurring and specifically, Convolutional Neural Networks (CNNs) and how they are able to be used to deblur images. It discusses the different forms of blur such as motion blur, out-of-focus blur, and mixed blur and compares these methods under the basis of blind and non-blind methods. The article sheds light on the various architecture and model design, loss functions, and performance indicators applied in image deblurring. Moreover, it draws attention to the issues that are presently observed in the sphere and gives possible path directions of the future research. The review has condensed and synthesized existing literature to provide a clear overview of the current solutions in image deblurring and offers guidance to the researchers on how to come up with the more precise, efficient, and adaptive methods of deblurring. The developments are meant to enhance the use of image restoration techniques in practical applications and this will lead to the quality and reliability of deblurring processes.
Cosmetic surgery is more prevalent in the world in recent years. A beautiful and flawless face is everyone's dream. Aging, environmental factors, disease, or poor diet are among the factors that influence body wrinkles. Various methods are used to reduce these lines. It can be said that the simplest and most effective solution is to inject cosmetic fluids into these areas. But, due to the increase in facial injections using cosmetic fluids, which are considered toxins, the risk of injury to the surrounding facial nerves and injury to one of the main facial nerves is increasing, creating a catastrophe or deformation in the face irreversibly. Deep learning algorithms have been used to determine whether cosmetic fluids are injected or not. Deep Convolutional Neural Networks (CNNs), VGG16, ResNet....etc deep learning algorithms have demonstrated excellent performance in terms of object detection, picture classification, and semantic segmentation. all the suggested approach consists of three stages: feature extraction, training, and testing/validation. Deep learning technology is used to train and test the system with before and after photographs. Numerous investigations have been carried out using various deep learning algorithms and databases the main goal is to attain maximum accuracy to ensure that injected cosmetic fluids by specialists have been injected in safe areas in addition to facial recognition and determining whether or not the person received an injection. The most used databases are IIITD plastic surgery and HDA_Plastic surgery.