Neural Network Light Source Prediction Model Adaptive Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Accurate estimation of light source information in image capture devices is challenging, especially in varying environments, which can distort color information and hinder correct diagnosis or image quality.
Innovation Solution
A method and system for establishing a light source information prediction model using a neural network trained with true light source information obtained from images of a white object, where the learning rate is adaptively adjusted based on predicted light source information to enhance training efficiency and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional white balance processing is used to estimate light source information, then the process is simple, but the accuracy of light source information estimation is insufficient leading to color distortion
Solution Approach 1:
The patent applies preliminary action by pre-training a neural network model offline using captured images and reference color information before actual white balance processing. The model is trained in advance with various lighting conditions and stored for later use, separating the complex training process from the real-time application, thus achieving high accuracy without increasing runtime system complexity
Solution Approach 2:
The patent introduces a neural network model as an intermediary between the captured image and the white balance correction process. This model acts as a mediator that processes the captured image data and reference color information to generate accurate light source information, replacing traditional direct estimation methods and improving measurement precision through learned patterns
2Productivity
If the learning rate is fixed during neural network training, then the training process is simple, but the training efficiency and accuracy are limited
Solution Approach 1:
The patent applies dynamics by implementing an adaptive learning rate adjustment mechanism where the learning rate changes dynamically during the training process based on training progress and performance metrics. This allows the training system to automatically optimize its own parameters, improving training efficiency and convergence without requiring manual intervention or complex external control systems
Data Source
AI summary
A method and a system for establishing a light source information prediction model are provided. A plurality of training images are captured for a target object. A white object is attached on the target object. True light source information of the training images is obtained according to a color of the white object in each of the training images. A neural network model is trained according to the training images and the true light source information, and a plurality of pieces of predicted light source information is generated according to the neural network model during the training. A learning rate for training the neural network model is adaptively adjusted based on the predicted light source information.