I was wondering since CNNs have dominated every image-related task. Is the Viola-Jones face detector still considered state-of-the-art, or have CNNs surpassed its performance?
According to the paper A Comparison of CNN-based Face and Head Detectors for Real-Time Video Surveillance Applications CNN based neural networks are significantly better than Viola-Jones in terms of precision but need more computational power to calculate the results. The mean average precision error for CNN based networks is 5 times less than for VJ for FDDB data. Here is the precision comparison table from the paper:
But VJ allows real time recognition with 60 frames per second. Here is the speed comparison from the same paper: