Target detection method based on lightweight network

Abstract

The invention provides a target detection method based on a lightweight network, and the target detection method comprises the steps: firstly obtaining a to-be-detected image, and obtaining a target processing value according to a convolutional neural network, wherein the target processing value comprises the resolution of the to-be-detected image and the depth and width of the convolutional neural network; performing composite optimization processing according to the target processing value to obtain the maximum accuracy (ACC) and the floating point operation rate (FLOPS) of the convolutionalneural network; optimizing the maximum accuracy (ACC) and a floating point preset operation rate (FLOPS) to obtain a target optimization value, the target optimization value being used for measuringthe detection efficiency of a basic network in the convolutional neural network; and finally, achieving target detection of the to-be-detected image by using the optimized convolutional neural network. The target extraction method has the beneficial effects that the network depth, the network width and the network resolution in the convolutional neural network are balanced, so that the picture output by the target extraction method is high in resolution and high in efficiency.

Description

technical field [0001] The present application relates to the field of computer technology, in particular to an effective target detection method in computer vision fields such as unmanned driving and assisted driving, and in particular to a target detection method based on a lightweight neural network. Background technique [0002] Object detection refers to the use of computer technology to detect and recognize the category and location information of objects of interest (such as vehicles, pedestrians, obstacles, etc.) in images or videos. It is one of the important research fields in the field of computer vision. With the continuous improvement and development of deep learning technology, object detection technology based on deep learning has a wide range of application scenarios in many real fields, such as: unmanned driving, assisted driving, face recognition, unmanned security, human-computer interaction , behavior recognition and other related fields. [0003] Convol...

AI Technical Summary

Problems solved by technology

See figure 1 , however, as the depth increases, the gradient corresponding to the network disappears, and the deeper network is more difficult to train

Although the current technology alleviates its training difficulty by adopting deeper networks: for example, ResNet-1000 (the 1000th layer of the ResNet network) has a deeper depth than ResNet-101, but the detection accuracy is similar

image 3 It shows the results of scaling the network resolution, which generally improves the accuracy of target detection, but for higher resolution input images, the detection accuracy still does not continue to increase when it reaches a certain level

[0009] In summary, it can be concluded that enlarging any dimension of network width, depth or resolution can improve accuracy, but for larger network architecture models, the output image target recognition rate will be reduced instead.

Images