Face detection technology detects the presence of a human face in an image and accurately frames its position. This technology is the basis of face matching, face verification, and face emotion analysis. It should have strong robustness and can adapt to the recognition challenges brought by illumination, occlusion, rotation, and color.
Early face recognition research mainly focused on face images with strong constraints. It is necessary to design clever face image textures and "features" of semantic expressions, such as HOG and Haar-like, to complete the training of recognition models.
With the enhancement of depth learning algorithm and GPU/FPGA computing power, an "end-to-end" face detection technology route has emerged. The learning of image features is integrated into the learning of neural network, and face detection, face key point detection and face image classification are output together. Obviously, the face detection method has entered a new stage and a new height.
This technology sharing comes from a new practice in the field of face recognition in the field of image processing projects, which has accumulated a certain experience of WIDER FACE and MTCNN, and welcomes technical exchanges among industry colleagues.
Lecturer: Datatang AI lab- Jin Zhenjie
At the Beijing Institute of Printing, he mainly studied face image classification, face key point detection, and face emotion analysis. Currently mainly engaged in the research and development of face recognition systems.
1.Overview of face and key detection
2.Face detection open source data set;
3.Face detection algorithm based on deep learning;
4.Report recent related practice work.
Highlights of this sharing: