Optimization in Makeup Recommendations Using RGB and YCbCr Analysis with CNN Algorithm for Skin Undertone

The objective of this inquiry is to extend the precision of a skin undertone-based beauty care products proposal framework by detecting skin photographs employing a blend of RGB colorspace, YCbCr, and the Convolutional Neural Arrange (CNN) calculation. In spite of the fact that the RGB colorspace has appeared to be exceedingly exact within the past, it is compelled by uneven light. To unravel this issue, this ponder proposes utilizing the YCbCr color plot, which is way better in tune with the human visual framework. The procedures incorporate picture increase, CNN show preparation, auxiliary information collecting through roboflow databases and overviews, and show execution evaluation utilizing measurements for exactness, exactness, review, and F1-score. The dataset used consists of 3000 vein images with a division of 80% for training data and 20% for testing data, which are classified into three classes: warm, neutral, and cool. The research results show that the CNN model with a combination of RGB and YCbCr colorspaces succeeded in achieving 99.83% accuracy in classifying skin undertones, improving the performance of the makeup recommendation system despite variations in lighting. These findings indicate that the proposed approach can provide more accurate and consistent makeup recommendations, thereby increasing customer satisfaction and loyalty and contributing to the fields of computer vision and machine learning.