Target category identification method, device and readable storage medium

Abstract

Embodiments of the present invention provide an object category identification method, device and readable storage medium. The method includes: extracting the correlation coefficient for each target category from the input image, multiplying each component in the target category correlation coefficient vector by a corresponding decimal, and subtracting each component in the obtained second correlation coefficient vector by Remove the maximum component, add the preset precision to each component in the obtained third correlation coefficient vector, and then perform rounding, update the components less than 0 in the obtained fourth correlation coefficient vector to 0, and calculate respectively 2 is the base number, and each component is used as the power obtained by the exponent, and the power corresponding to each component in the fifth correlation coefficient vector is divided by the sum of the powers corresponding to all components in the fifth correlation coefficient vector, and each quotient of the obtained The values ​​are the probabilities of each type of target contained in the input image. The embodiments of the present invention reduce the computational complexity of target category recognition on the premise of ensuring the target category recognition accuracy.

Description

technical field [0001] The invention relates to the technical field of data processing, in particular to a method, a device and a readable storage medium for identifying an object category. Background technique [0002] With the rapid development of deep learning technology, deep learning algorithms have been widely used in the field of artificial intelligence. Classification is the most common application scenario, such as expression recognition, age recognition, gender recognition, scene recognition, sound event classification, and so on. Due to the very good classification effect and relatively simple calculation process of softmax, softmax classifiers are often used in various classification tasks. [0003] The calculation process of softmax requires multiple exponential operations with the natural number e as the base. This exponential operation is not easy to implement on hardware such as NPU (Neural-Network Processing Unit, neural network processing unit), which affe...

AI Technical Summary

Problems solved by technology

[0003] The calculation process of softmax requires multiple exponential operations with the natural number e as the base. This exponential operation is not easy to implement on hardware such as NPU (Neural-Network Processing Unit, neural network processing unit), which affects the algorithm model containing the softmax classifier. Use on hardware such as NPU

Images