Target detection model optimization and acceleration method for few types of scenes

Abstract

The invention relates to the technical field of computer vision, in particular to a target detection model optimization and acceleration method for few types of scenes, which comprises the following steps of: obtaining and labeling a to-be-detected picture; adding the marked picture into a feature extraction device for feature extraction to obtain three groups of feature maps with fixed sizes; sending the three groups of feature maps with the fixed sizes into a prediction device with a FocalLoss loss function for result prediction; carrying out detection frame filtering on a prediction resultsuch that a same to-be-detected object can be detected, wherein only one detection frame is outputted. The feature extraction process comprises the following steps of: preprocessing DarkNet; alexNet,the method comprises the steps of compressing networks such as ResNet, VGG, GoogLeNet, SENet and DenseNet, adjusting the size of a picture to be N * N, enabling the selectable value of N to be 32 times of 320-1280, and performing feature extraction on the picture by adopting a compressed feature extraction network. According to the invention, high detection speed and high accuracy can be achievedat the same time.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a method for optimizing and accelerating an object detection model in a scene with few categories. Background technique [0002] The target detection task is an important application in the field of computer vision. Its main goal is to find all target objects from the image and give their classification information and location information. [0003] Based on traditional pattern recognition methods, it usually takes tens of seconds to process a picture, while the target detection algorithm based on convolutional neural network can already complete the target detection of a high-resolution picture in hundreds or even tens of milliseconds. Work, but also has a great improvement in accuracy. [0004] The rapid development of neural networks has made people move away from the method of manually designing features, and turned to the method of training deep convolutional neural...

AI Technical Summary

Problems solved by technology

In the second stage, the convolutional neural network further corrects the coordinate position of the region of interest extracted in the first stage, and after passing all the prediction results through the non-maximum value suppression algorithm, the final result is output, and the object in the picture is obtained. Precise position and coordinates, it can be seen that this type of method is characterized by more accurate results, but its corresponding time overhead is also relatively large

[0007] The second category: single-stage detectors, such as YOLO, SSD, its main feature is that it only needs one step to directly obtain the location information and classification information of the object, different from the two-stage detector, the network is directly in the feature map of the last layer Output the final prediction result. For each point of the last layer feature map, output the position offset relative to the preset anchor point, and directly obtain the position information of each object. Similarly, the final result also needs to pass Non-maximum suppression algorithm, since this type of method has only one stage, it is much faster than the two-stage detector, but the accuracy is relatively low

[0008] The third category: methods without anchor points. The characteristic of this type of method is that it does not need to set anchor points in advance, and directly regresses the four boundaries of each object. At present, the accuracy can reach a level equivalent to that of a two-stage detector. But the speed is generally slow and the practicability is poor

Images