US 11,869,224 B2
Method and system for establishing light source information prediction model
Yi-Jin Huang, New Taipei (TW); Chien-Hung Li, New Taipei (TW); and Yin-Hsong Hsu, New Taipei (TW)
Assigned to Acer Incorporated, New Taipei (TW)
Filed by Acer Incorporated, New Taipei (TW)
Filed on Oct. 4, 2022, as Appl. No. 17/960,108.
Application 17/960,108 is a continuation of application No. 16/813,741, filed on Mar. 10, 2020, granted, now 11,494,585.
Claims priority of application No. 109101420 (TW), filed on Jan. 15, 2020.
Prior Publication US 2023/0045128 A1, Feb. 9, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G06K 9/62 (2022.01); G06V 10/141 (2022.01); G06V 10/56 (2022.01); G06F 18/214 (2023.01); G06F 18/21 (2023.01); G06V 10/764 (2022.01); G06V 10/774 (2022.01); G06V 10/776 (2022.01); G06V 10/82 (2022.01)
CPC G06V 10/141 (2022.01) [G06F 18/214 (2023.01); G06F 18/217 (2023.01); G06V 10/56 (2022.01); G06V 10/764 (2022.01); G06V 10/774 (2022.01); G06V 10/776 (2022.01); G06V 10/82 (2022.01)] 9 Claims
OG exemplary drawing
 
1. A method for establishing a light source information prediction model, comprising:
capturing a plurality of training images for a target object, wherein a white object is attached on the target object;
obtaining true light source information of the training images according to a color of the white object in each of the training images; and
training a neural network model according to the training images and the true light source information, and generating a plurality of pieces of predicted light source information according to the neural network model during the training, wherein
a learning rate for training the neural network model is adaptively adjusted based on the predicted light source information;
wherein training the neural network model according to the training images and the true light source information and generating the pieces of predicted light source information according to the neural network model during the training further comprises:
performing white balance processing on the training images according to the predicted light source information to obtain a plurality of adjustment images;
determining the learning rate for training the neural network model according to the adjustment images;
inputting the predicted light source information and the corresponding true light source information into a loss function to generate a loss value;
wherein a weight information be adjusted in the neural network model according to the loss value.