Image target detection method and system, electronic equipment and storage medium

Abstract

The invention discloses an image target detection method, which is applied to a parallel processing chip of a heterogeneous platform, and comprises the steps of receiving bounding box information of aplurality of target object bounding boxes obtained by performing target identification on an image through a convolutional neural network; storing the bounding box information to a random access memory RAM; executing a local maximum search operation based on a non-maximum suppression algorithm on the read bounding box information in a multi-path parallel manner to obtain a preset number of to-be-output bounding boxes; and transmitting the bounding box information of the bounding box to be output to a central processing unit (CPU) of the heterogeneous platform, so that the CPU outputs an imagetarget detection result. According to the invention, the communication data volume of the parallel processing chip and the CPU in the image target detection process can be reduced, and the image target detection efficiency is improved. The invention further discloses an image target detection system, electronic equipment and a storage medium, which have the above beneficial effects.

Description

technical field [0001] The present application relates to the field of computer technology, in particular to an image target detection method, system, electronic equipment and storage medium. Background technique [0002] In the process of image object detection, it is necessary to generate bounding boxes according to the types of objects. For bounding boxes of the same category, NMS (Non-Maximum Suppression, non-maximum suppression algorithm) calculation needs to sort the scores of all bounding boxes in this category, first select the highest score and its corresponding box, and then traverse the remaining boxes , if the overlapping area with the current highest sub-frame is greater than a certain threshold, the frame will be deleted. Then continue to select a bounding box with the highest score from the next category, repeat the above process, and finally get the detection results of all categories. [0003] In related technologies, convolution calculation is first perfo...

AI Technical Summary

Problems solved by technology

However, due to the characteristics of serial computing, the CPU takes a long time to complete the NMS calculation.

When using the above-mentioned related technologies to realize image target detection, it is necessary to receive a large amount of data transmitted by FPGA and perform serial processing on the data, so that the efficiency of target detection is low, and NMS calculation becomes the calculation bottleneck of target detection.

Images