Method and apparatus for real-time eye tracking of kiosk userand by performing inference on calibration model
Proposed are a method and apparatus including real-time eye tracking and calibration functions in a kiosk including a display. The method may include collecting first data related to a user's appearance based on image data on the user acquired from a camera of the kiosk, and generating first screen coordinates of the display corresponding to the user's eyes based on the first data and performing eye tracking on the user. The method may also include collecting second data including second screen coordinates for a corresponding touch point detected when the user touches and selects a menu displayed on the display and training a calibration machine learning model using the collected second data, and performing inference on the calibration model using the first screen coordinates according to the eye tracking performed at a time other than the touch selection time point to perform calibration on the eye tracking.
1 . A method performed by a kiosk including a display having a touch screen, comprising:
collecting first data related to a user's appearance based on image data on the user acquired from a camera of the kiosk;
generating first screen coordinates of the display corresponding to the user's eyes based on the first data and performing eye tracking on the user;
collecting second data including second screen coordinates for a corresponding touch point detected when the user touches and selects a menu displayed on the display and training a calibration model, which is a machine learning model, using the collected second data; and
performing inference on the calibration model using the first screen coordinates according to the eye tracking performed at a time other than a touch selection time point to perform calibration on the eye tracking,
wherein the training of the calibration model includes collecting the second data according to touch selection performed upon a required action in which the user purchases a ticket or places an order using the kiosk,
wherein the required action includes an initial selection action in which the user who intends to use the kiosk initially selects an initial menu displayed on the display, and an intermediate selection action in which the user selects detailed menus displayed on the display to select detailed items of a desired ticket or order after the initial selection action, and
wherein the training of the calibration model includes gradually training the calibration model using a plurality of pieces of second data collected according to touch selections on a plurality of menus.
2 . The method of claim 1 , wherein the first data includes data on a position, facial pose, and distance of the user.
3 . The method of claim 1 , wherein the performing of the eye tracking is performed using an eye tracking model, which is a machine learning model, and
when the first data is input, the eye tracking model is a model trained to output the first screen coordinates corresponding to the first data.
4 . The method of claim 1 , wherein, when the first screen coordinates are input, the calibration model is a model trained to output the second screen coordinates, which are a result of calibrating the first screen coordinates.
5 . The method of claim 1 , further comprising providing a customized interface for the user through the display using the calibrated second screen coordinates.
6 . The method of claim 5 , wherein the customized interface provides functions of analyzing eye information of the user according to the calibrated second screen coordinates and adjusting or recommending a screen interface.
7 . An apparatus related to a kiosk, comprising:
a camera configured to acquire image data on a user of the kiosk;
a display comprising a touch screen; and
a controller configured to provide a function of the kiosk using the image data,
wherein the controller is configured to:
collect first data related to the user's appearance based on the image data;
generate first screen coordinates of the display corresponding to the user's eyes based on the first data and perform eye tracking on the user;
collect second data including second screen coordinates for a corresponding touch point detected when the user touches and selects a menu displayed on the display and train a calibration model, which is a machine learning model, using the collected second data; and
perform inference on the calibration model using the first screen coordinates according to the eye tracking performed at a time other than a touch selection time point to perform calibration on the eye tracking,
wherein the controller is configured to collect the second data according to touch selection performed upon a required action in which the user purchases a ticket or places an order using the kiosk when performing the calibration,
wherein the required action includes an initial selection action in which the user who intends to use the kiosk initially selects an initial menu displayed on the display and an intermediate selection action in which the user selects detailed menus displayed on the display to select detailed items of a desired ticket or order after the initial selection action, and
wherein the controller is configured to gradually train the calibration model using a plurality of pieces of second data collected according to touch selections for a plurality of menus in the initial selection action and the intermediate selection action.
8 . The apparatus of claim 7 , wherein the controller is configured to use the first data including data on a position, facial pose, and distance of the user.
9 . The apparatus of claim 7 , wherein, when receiving the first data, the controller is configured to perform the eye tracking using an eye tracking model, which is a machine learning model trained to output the first screen coordinates corresponding to the first data.
10 . The apparatus of claim 7 , wherein, when receiving the first screen coordinates, the controller is configured to use the calibration model trained to output the second screen coordinates, which are a result of calibrating the first screen coordinates.
11 . The apparatus of claim 7 , wherein the controller is configured to provide a customized interface for the user through the display using the calibrated second screen coordinates.
12 . The apparatus of claim 11 , wherein the controller is configured to provide the customized interface by providing functions of analyzing eye information of the user according to the calibrated second screen coordinates and adjusting or recommending a screen interface.