Fast object detection method based on convolutional neural network

Abstract

A fast object detection method based on convolutional neural network, involving computer vision technology. First, use the training set to train the parameters of the convolutional neural network, and then use the extended graph to solve the problem of the maximum pooling loss feature and generate a discriminative complete feature map; regard the fully connected weight of the convolutional neural network as a linear classifier, and use possible Approximate learning framework to estimate the generalization error of the linear classifier on the discriminative complete feature; estimate the number of required linear classifiers according to the generalization error and the expected generalization error threshold, and finally use the linear classifier on the discriminative complete feature map Target detection is done based on smoothing window. Significantly improve detection efficiency and target detection accuracy.

Description

technical field [0001] The invention relates to computer vision technology, in particular to a fast target detection method based on a convolutional neural network. Background technique [0002] An important source of human perception of the world is through image information. Studies have shown that about 80% to 90% of the information that humans obtain from the outside world comes from image information obtained by human eyes. Human beings have a high ability to perceive external image information, and can quickly locate and analyze targets. In order for a computer to have powerful visual perception and understanding capabilities, it should have powerful target detection and recognition capabilities similar to humans. Object detection is an important prerequisite for visual perception and object understanding. The efficiency and accuracy of object acquisition determine the speed and effect of visual perception. Once the computer has the powerful target detection and perc...

AI Technical Summary

Problems solved by technology

However, this method of separately performing the classification of the convolutional neural network and the target detection in the deformation model makes the detection effect of the overall framework not very outstanding, and the efficiency of this model is not very high.

Images