The Effect of Disguise and Makeup in Facial Recognition

Abstract

Recently, facial recognition technologies have seen significant advancements. However, their performance degrades when faces are altered with disguises and makeup. Improving the robustness of these technologies is crucial due to their wide range of applications that require high accuracy. By studying the possible variations, this paper aims to evaluate the performance of hand-crafted and deep learning methods in identifying faces affected by these challenges. The study is conducted on two challenging datasets: Disguise Faces and Makeup dataset (DMFD) and Celebrity Frontal Profile dataset (CFP). The handcrafted methods employed in this study include well-known feature extraction techniques such as Principal Component Analysis (PCA), Local Binary Patterns (LBP), Scale-Invariant Feature Transform (SIFT), and Speeded-Up Robust Features (SURF). Furthermore, the study demonstrates the use of deep learning approaches, including ResNet and EfficientNet, which are widely used for classification tasks. The effectiveness of these approaches is detailed to offer an in-depth comparison of the accuracy achieved by traditional hand-crafted methods versus deep learning techniques. This comparative analysis can help identify various challenges and inspire new ideas for enhancing the overall performance of facial recognition systems © 2025 IEEE.