Learning image processing method, system and server

Abstract

The embodiment of the present invention discloses a learning-type image processing method, system and server, including the following steps: collecting the target image to be measured; inputting the target image to be measured into a preset convolutional neural network model, and obtaining the The convolutional neural network model outputs classification data in response to the input of the face image, and the convolutional neural network model takes a loss function as a constraint condition, limiting the cosine distance of the features in the classification data to tend to the Euclidean distance; Acquiring the classification data, and performing content understanding on the target image to be tested according to the classification data. The cosine distance in the classification data is maximized by filtering the classification data through the loss function based on the cosine distance in the joint loss function. However, due to the single color in the simple image, the convergence within the class is strong. The maximization of the cosine distance, On the contrary, the cosine distance tends to be the calculation result of the Euclidean distance, thereby simplifying the complexity of the implementation.

Description

technical field [0001] Embodiments of the present invention relate to the field of image processing, in particular to a learning image processing method, system and server. Background technique [0002] With the development of deep learning technology, the convolutional neural network has become a powerful tool for extracting face features. For the convolutional neural network with a fixed model, the core technology is how to design the loss function so that it can effectively supervise the convolutional neural network. Convolutional neural network training, so that the convolutional neural network has the ability to extract face features. [0003] The loss function commonly used in the prior art based on Softmax+Centerloss is the most commonly used supervised loss function because of its good effect and simple training. The Centerloss loss function refers to counting an average center point for each class, and then using the The Euclidean distance between each sample and t...

Images