Living organism recognition neural network of RGB monocular image and construction method thereof

Abstract

The invention discloses a living organism recognition neural network of an RGB monocular image. The living organism recognition neural network comprises a root module, a plurality of repeatable modules, a feature extraction module and an output module which are sequentially connected from front to back, wherein the root module comprises a convolution layer, a batch normalization layer and an activation layer which are sequentially connected and packaged from front to back; wherein the repeatable module comprises a convolution layer, a batch normalization layer, an activation layer, a depth separable convolution layer, a batch normalization layer, an activation layer, a convolution layer and a batch normalization layer which are sequentially connected and packaged from front to back; if therepeatable module performs down-sampling, a spatial attention layer is further arranged at the tail end of the repeatable module; the feature extraction module comprises a global average pooling layer, a full connection layer, an activation layer and a regularization layer which are sequentially connected and packaged from front to back; and the output module is a full connection layer of which the weight value is subjected to regularization processing. According to the scheme, the living organism recognition neural network has the advantages of simple logic, low technical workload, high calculation precision and the like.

Description

technical field [0001] The invention relates to the technical field of biological living body recognition in computer face recognition, in particular to a biological living body recognition neural network of RGB monocular images and a construction method thereof. Background technique [0002] In the field of biological living body recognition technology in computer face recognition, computer face recognition technology uses cameras or sensors to collect relevant information such as face images, and performs functions such as identity comparison, identity confirmation, and attribute identification. At present, computer face recognition technology has been widely used in many fields such as security, attendance, finance, transportation, and smart terminals. In practical applications, it is necessary to accurately judge whether the face collected by the camera or sensor is from a living organism, that is, a real natural person, rather than an attack by a non-living organism, su...

AI Technical Summary

Problems solved by technology

For some applications, such as some mobile Internet applications that need to use the front camera of the mobile phone for authentication, these technologies cannot be used

[0007] As we all know, because the information provided by monocular RGB images is more limited than other methods such as 3D structured light, and in the actual application process, the environment, lighting, attack methods of non-biological living objects, and the specifications and models of the acquisition cameras are all different. There are many different possibilities; it can be seen that the neural network trained with monocular RGB images leads to low accuracy and is prone to over-fitting problems, that is, it has better performance in some scenes, but not in others. In the scenario, the performance of the network drops significantly

Images