Target detection method, device, terminal device and computer-readable storage medium

Abstract

The invention discloses a target detection method, which relates to the artificial neural network field. The method comprises the following steps: inputting an image to be detected into a target detection network to obtain a target confidence diagram marked with the confidence level of each pixel of the image to be detected; Wherein the confidence level of each pixel refers to the confidence levelof each pixel as an integral part of the target area; selecting a target pixel whose confidence level conforms to a preset standard from the target confidence diagram, and judging a region of a preset shape centered on the target pixel to be a target region of the image to be detected. In addition, the invention also discloses a target detection device, a terminal device and a storage medium, which can effectively improve the reliability of the target detection result, simultaneously reduce the complexity of the target detection algorithm and improve the efficiency of the target detection.

Description

technical field [0001] The present invention relates to the field of artificial neural networks, in particular to a target detection method, device, terminal equipment and computer-readable storage medium. Background technique [0002] Target detection refers to detecting the target area where the target object is located from a given image. Face-related tasks such as face recognition, face beauty makeup, and age estimation that are popular nowadays all require target detection including face detection as a prerequisite. [0003] In the prior art, a binary classification method is usually used to complete the target detection task. The image to be detected is detected by a plurality of binary classifiers, wherein the first binary classifier detects a large number of suspected target areas from the image to be detected, and the subsequent multiple binary classifiers sequentially detect the large number of suspected target areas. The target area is subjected to multiple bina...

AI Technical Summary

Problems solved by technology

[0005] 1. The binary classification method is used to process the target detection task, which is prone to ambiguity and cannot make a reliable judgment when the target image is incomplete. For example, in the case of face detection, part of the image of the face is blocked, etc., the same binary classification is used Method to detect the same image to be detected may get different results;

[0006] 2. Due to the need for multiple binary classifiers to detect the image to be detected, the complexity of the algorithm is high;

[0007] 3. Since the binary classification method cannot determine the exact range of the target area, an additional regression network is needed to return the window where the target area is located, which further increases the complexity of the algorithm

Images