In this paper, we propose a novel approach to identifying celebrity cartoon faces by integrating conventional (HoG) and convolutional feature (FaceNet) extraction techniques for recognition of celebrity cartoon faces. The features are analyzed by combining them in different ways, including a single feature, fusing two features, or fusing all three features at once. Furthermore, distinct conventional classifiers are used to categorize into 100 distinct classes by varying hyperparameters. Experimentally, it is evident that an outstanding recognition rate is achieved by the fusion of FaceNet features with HoG. It is observed that the SVM (RBF kernel) is the most suitable learning model for recognition. The effectiveness of our model is tested on a standard dataset, namely Cartoon Faces in the Wild (IIIT-CFW), which includes 8928 cartoon faces of 100 different celebrities. The classification results are validated using the F-measure as a classification metric. From our experiments, we found that feature fusion yields outstanding results. Based on a quantitative analysis of our model, achieve the state-of-the-art F-measure of 80.23% recognition rate using 100 distinct classes of celebrity cartoon face images.