“…To aid the performance of robust image watermarking, modern classifiers were, recently, tested. This includes models like Particle Swarm Optimization (PSO) [14], Support Vector Machines (SVM) [15], Random Forest (RF) [16], and Artificial Neural Networks (ANNs) and deep learning approaches [17][18][19][20][21][22][23][24][25][26][27][28][29][30][31]. Detailed survey on these classifiers has shown that: a) PSO exhibited relatively high robustness with poor performance against cropping and JPEG compression; b) SVM experienced complications especially when choosing the proper kernel, and when applying the finetune parameter regularization; c) RF has revealed difficulties when training untested data with increasing complexity raised at increased data size, and, d) ANN exhibited better performance over PSO, SVM and RF, since it can be easily learned, even with complex patterns and large dataset sizes.…”