Method and system for analyzing hair characteristics
AssigneeMacroHI Co., Ltd.
InventorsSin-Ye Jhong, Chih-Hsien Hsia, Chun-Wei Chen
Life of this patent
Estimating the term…
Abstract
A hair feature analysis method and system are provided. The hair feature analysis method includes the steps of capturing high magnification and low magnification images of a scalp area and performing preprocessing on these images. The preprocessed high magnification and low magnification images are then input into an artificial intelligence model to simultaneously detect hair follicles and calculate hair widths in the scalp area. Hair characteristics are then calculated based on the analyzed high magnification and low magnification images. The artificial intelligence model can be an R-CNN model or a variant thereof.
Claims10
A hair feature analysis method, comprising:
- capturing a high magnification image and a low magnification image of a scalp area, providing detailed information about follicles by the high magnification image, providing an overall view of hair density and distribution by the low magnification image;
- preprocessing the high magnification image and low magnification image, wherein the step of preprocessing the captured high magnification image and low magnification image includes denoising, resizing, normalizing, or recapturing the image;
- inputting the preprocessed high magnification image and low magnification image into an artificial intelligence model, simultaneously detecting multiple hair follicles and multiple hair widths in the scalp area;
- calculating at least one hair feature from the high magnification image and the low magnification image analyzed by the artificial intelligence model, wherein the hair feature includes average hair diameter, thin hair density ratio, or total hair density ratio;
- and
- wherein, the artificial intelligence model is a Region-based Convolutional Neural Network (R-CNN) variant in the form of an HTC enhanced model that is modified from a hybrid task cascade model, and the HTC enhanced model is formed by:
- replacing a semantic segmentation module in the hybrid task cascade model with a global feature enhancement module;
- connecting refined semantic features from the global feature enhancement module with multiple framework branches;
- and
- introducing at least one framework-mask enhancement module, connecting the framework branches and at least one mask branch.
A hair feature analysis system, comprising:
- an image capturing device for capturing a high magnification image and a low magnification image of a scalp area, wherein the high magnification image provides detailed information about follicles, and the low magnification image provides an overall view of hair density and distribution;
- a preprocessing module for preprocessing the captured high magnification image and low magnification image, wherein the step of preprocessing the captured high magnification image and low magnification image includes denoising, resizing, normalizing, or recapturing the image;
- an artificial intelligence model for simultaneously detecting multiple hair follicles and calculating multiple hair widths in the high magnification image and low magnification image, wherein the artificial intelligence model is a variant of a Region-based Convolutional Neural Network (R-CNN) model;
- and
- an analysis module for receiving the high magnification image and low magnification image analyzed by the artificial intelligence model and calculating at least one hair feature, wherein the hair feature includes average hair diameter, thin hair density ratio, or total hair density ratio;
- wherein the artificial intelligence model is an HTC enhanced model, which is modified from a hybrid task cascade model, and the HTC enhanced model is formed by:
- replacing a semantic segmentation module in the hybrid task cascade model with a global feature enhancement module;
- connecting refined semantic features from the global feature enhancement module with multiple framework branches;
- and
- introducing at least one framework-mask enhancement module, connecting the framework branches and at least one mask branch.
Dependencies are inferred from the claim text.