Face detection method based on cascaded depth network

Abstract

The invention discloses a face key point detection method based on a cascade depth network, comprising the following steps: sample collection: face information, eye information, nose information and mouth information are respectively collected in advance; the position of the face characteristic information is marked to determine the eye key point information, nose tip key point information and mouth corner key point information; the face key point detection method comprises the following steps: the face information, the eye information, the nose information and the mouth corner key point information are collected in advance; First model training: the collected face information, eye information, nose information and mouth information will be trained separately the first part of the network.As that first part network and the second part network are use for cooperating with the first model train and the second model training, It not only can better locate the key points of human face, but also has high precision, can adapt to the application needs in different environments, improve the anti-interference ability and solve the deviation of the key points of human face positioning, andface occlusion, lighting and make-up sampling difficult problems.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a method for detecting human face key points based on a cascaded deep network. Background technique [0002] Face key point positioning has many applications in the field of computer vision, such as face recognition, expression recognition, animation production, etc. In recent years, face key point recognition has made great progress, but under the influence of occlusion, lighting, makeup, etc., there are still big problems. At present, face key point localization mainly includes model-based methods, regression-based methods, and deep learning-based methods. [0003] Model-based methods mainly use local features around key points to construct face models during training. This method does not require a large amount of data to train the model, and the obtained parameterized model is highly versatile and can be used in different application scenarios such as detection an...

AI Technical Summary

Problems solved by technology

In addition, when fitting the model, it often involves solving optimization problems, so the prediction speed is also slow

Images