Dog face detection method and device

Abstract

The embodiment of the invention discloses a dog face detection method and device. The method comprises the steps: carrying out normalization of an image, changing the size of the image, and transmitting the converted data into a backbone network for feature extraction; then sending a feature extraction result to an area generation module to generate a more accurate target object bounding box; performing target generation area pooling on the candidate target area and the feature map of the backbone network, sending to a predictor after passing through a flattening layer and two full-connection layers, calculating loss, filtering out a low-score detection object by using a maximum suppression method (NMS), and finally, performing generalized transformation on a prediction frame and an original image correspond to each other to form a detected image. The detection accuracy is very high for the pictures with multiple dogs in one picture, different sizes of dogs in the picture or very complex scenes, and the technical problem of low detection accuracy for the pictures and dog faces with the complex scenes in the related technology is solved.

Description

technical field [0001] The present disclosure relates to the technical field of data processing, and in particular to a dog face detection method and device. Background technique [0002] With the development of the national economy and people's lives, the pet dog industry has formed a series of complete product and service chains, and security issues such as stray dogs hurting people are also constantly appearing around people's lives. Therefore, urban public management and interpersonal issues caused by dogs Relationship issues have become a major problem that needs to be solved urgently. [0003] Aiming at the problem of dog security management, dog face detection technology based on machine learning has emerged as the times require. Compared with traditional methods such as ID tags and manual chip injection, it has many advantages in terms of efficiency, reliability, cost and operability. In the process of traditional HOG feature extraction, the computational complexity...

AI Technical Summary

Problems solved by technology

[0004] However, in dog face detection, the data has dog face scale problems and many small samples, and it is impossible to accurately identify dog ​​faces from pictures with complex environments.

For example, in the practical application of dog face detection, the received dog face data often contains many other interfering objects, such as people, surrounding environment objects, and for stray dogs, their environment is more complex

Images