Convolutional network visualization method and device for image recognition

Abstract

The invention discloses a convolutional network visualization method and device for image recognition, and relates to the field of artificial intelligence deep learning visualization. The convolutional network visualization method includes the steps: loading a ubuntu mirror image by utilizing a Docker container; installing a Caffe on the basis of docker; utilizing a Docker container to package a front-end deep learning visualization tool box; based on a Caffe network architecture, utilizing a corresponding convolutional neural network model to identify an image, and displaying a convolution kernel of the convolutional neural network model in an image identification process, an image change formed after convolution of an original image and corresponding parameters layer by layer through a visualization interface of a deep learning visualization toolbox; and by utilizing a convolutional neural network visualization technology, enabling the whole image recognition process to be subjectedto imaging visual interpretation, so that more accurate judgment on an image is facilitated, and meanwhile, a more persuasive basis is provided for a result.

Description

technical field [0001] The invention discloses a visualization method and device, and relates to the field of artificial intelligence deep learning visualization, in particular to a convolutional network visualization method and device for image recognition. Background technique [0002] As one of the key technologies involved in big data analysis, deep learning overcomes the limitations of traditional machine learning algorithms relying on artificial feature establishment and screening, and has achieved good results in many fields such as speech recognition, visual object recognition, target detection, and drug discovery. practical results. The "feature learning" represented by deep learning enables the computer to automatically find the high-dimensional related feature values ​​of the target based on big data, establish a data processing channel model, realize a fully automatic intelligent processing process, and complete the specified application scenarios. Target detect...

AI Technical Summary

Problems solved by technology

[0003] The convolutional neural network in deep learning is often used for image recognition, model building, and effective recognition. However, the convolution calculation process and recognition results do not yet have perfect detection and tracking and intuitive display and interpretation methods, resulting in convolutional neural networks. The decisions made by the product neural network are often difficult to be accepted by the public, so that the application cannot be fully promoted

Images