Evaluation method and evaluation device for facial key point positioning result
AssigneeTencent Technology (Shenzhen) Company Limited
InventorsChengjie Wang, Jilin Li, Feiyue Huang, Kekai Sheng, Weiming Dong
Life of this patent
Estimating the term…
Abstract
Disclosed are an evaluation method and an evaluation device for a facial key point positioning result. In some embodiments, the evaluation method includes: acquiring a facial image and one or more positioning result coordinates of a key point of the facial image; performing a normalization process on the positioning result coordinate and an average facial model to obtain a normalized facial image; and extracting a facial feature value of the normalized facial image and calculating an evaluation result based on the facial feature value and a weight vector.
Claims17
An evaluation method for a facial key point positioning result, comprising:
- by a terminal device for performing facial image analysis, the terminal device including one or more processors and memory storing instructions executed by the processors:
- acquiring one or more facial images and key point coordinates marked manually;
- performing a positioning calculation on the one or more facial images using an Explicit Shape Regressor (ESR) algorithm to obtain one or more key point coordinates;
- calculating a Root of Mean Squared Error (RMSE) between the one or more key point coordinates and the key point coordinates marked manually;
- determining an accuracy score of the one or more key point coordinates based on the calculated RMSE, wherein the greater the RMSE, the less the accuracy score;
- reducing the one or more key point coordinates to the average facial model to obtain a reference facial image, and extracting a facial feature value of the reference facial image;
- calculating a weight vector based on the facial feature value and the accuracy score;
- and
- by the terminal device or another terminal device,
- evaluating a facial key point positioning result based on the weight vector.
An evaluation device for a facial key point positioning result, applied in a terminal device for performing facial image analysis, and comprising a processor and a memory for storing program instructions, wherein the processor executes the program instructions to:
- acquire one or more facial images and key point coordinates marked manually;
- perform a positioning calculation on the one or more facial images by using an Explicit Shape Regressor (ESR) algorithm to obtain one or more key point coordinates;
- calculate a Root of Mean Squared Error (RMSE) between the one or more key point coordinates and the key point coordinates marked manually;
- determine an accuracy score of the one or more key point coordinates based on the calculated RMSE, wherein the greater the RMSE, the less the accuracy score;
- reduce the one or more key point coordinates to the average facial model to obtain a reference facial image, and extracting a facial feature value of the reference facial image;
- calculate a weight vector based on the facial feature value and the accuracy score;
- and
- evaluate a facial key point positioning result based on the weight vector.
A non-transitory computer storage medium, applied in a terminal device for performing facial image analysis, comprising computer executable instructions, wherein the computer executable instructions are adapted to perform an evaluation method for a facial key point positioning result, wherein the evaluation method for a facial key point positioning result comprises:
- acquiring one or more facial images and key point coordinates marked manually;
- performing a positioning calculation on the one or more facial images using an Explicit Shape Regressor (ESR) algorithm to obtain one or more key point coordinates;
- calculating a Root of Mean Squared Error (RMSE) between the one or more key point coordinates and the key point coordinates marked manually;
- determining an accuracy score of the one or more key point coordinates based on the calculated RMSE, wherein the greater the RMSE, the less the accuracy score;
- reducing the one or more key point coordinates to the average facial model to obtain a reference facial image, and extracting a facial feature value of the reference facial image;
- calculating a weight vector based on the facial feature value and the accuracy score;
- and
- evaluating a facial key point positioning result based on the weight vector.
Dependencies are inferred from the claim text.