A method for generating an alignment model of a target object

Abstract

The invention discloses a method for generating an alignment model of a target object, which is suitable for execution in a computing device. After the image is subjected to data enhancement processing, the first training image is obtained; the first training image is exposed to obtain the second training image; and the first training image and the second training image are input into the alignment model of the pre-trained target object for training, to generate an alignment model of the trained target object. This solution can improve the accuracy of the model, effectively improve the problem of unstable key point alignment, and at the same time enable the model to meet the application requirements of the mobile terminal.

Description

technical field [0001] The present invention relates to the technical field, in particular to a method for generating an alignment model of a target object, a method for aligning a target object, a computing device, and a storage medium. Background technique [0002] Alignment of key points or feature points of target objects such as cat faces and dog faces in images is widely used in many real-world scenarios. For example, in the process of taking a photo or group photo of a pet cat, by detecting the facial features and outline points of the cat, some texture controls or text can be added in real time to increase the fun of shooting. However, unlike human face alignment, pets are lively and active, resulting in various postures and expressions, and the variety of pets increases the difficulty of object feature point alignment. [0003] The feature point alignment method based on the mobile terminal requires high real-time performance. The current alignment model based on ...

AI Technical Summary

Problems solved by technology

At present, the alignment model based on the convolutional neural network is relatively large and cannot be deployed on the mobile terminal; and the calculation efficiency of the model is low, and the key points cannot be detected in real time, and there is a phenomenon of key point jitter instability during the alignment process

[0004] The existing cat face or dog face alignment methods are mainly divided into two categories. One is to detect directly based on the face point detection model of the convolutional neural network. The accuracy is low, which affects the positioning and addition of subsequent controls.

One is to use a more complex convolutional neural network for deep learning. Although it can achieve higher accuracy, it has a large amount of calculation and low efficiency and cannot adapt to real-time requirements.

[0005] Therefore, there is a need for an alignment model of the target object, which can match the computing efficiency and memory of the mobile terminal, and improve key point jitter and network training instability.

Images