MTCNN: 3 CNNs Outperforming One Zhuge Liang
Introduction MTCNN (Multi-Task Cascaded Convolutional Networks) was proposed in 2016. As the name suggests, the algorithm adopts a cascaded CNN structure to achieve multi-task learning—face detection and face alignment, ultimately predicting the positions of face bounding boxes and landmarks. MTCNN achieved the best results at that time (April 2016) on the FDDB, WIDER FACE, and … Read more