Adaptive Object Verification Model Updating via Segmented Image Sets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Object verification models based on machine learning face challenges in maintaining accuracy when new features of the target object appear, as these features were not captured during the initial training process, leading to reduced verification efficiency and accuracy.
Innovation Solution
A method for adaptive model updating is proposed, where a first small image set and its annotation information are used to update the model, and if the accuracy falls below a threshold, a larger image set with corresponding annotation information is employed to further refine the model, balancing computing overhead and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the model is updated frequently using large image sets to maintain accuracy, then verification accuracy is improved, but computing overhead and retraining time increase significantly
Solution Approach 1:
The patent segments the model updating process into two distinct phases: frequent updates using small image sets to capture new features quickly, and periodic updates using large image sets to maintain overall accuracy. This segmentation allows the system to balance between responsiveness to new features and maintenance of high accuracy without requiring frequent large-scale retraining
Solution Approach 2:
The patent implements periodic action by scheduling large image set updates at intervals rather than continuously. The system performs frequent small updates as needed, then periodically performs comprehensive updates using large image sets to refresh the model's overall knowledge base, reducing total retraining time while maintaining accuracy
2Reliability
If a large image set is used for model updating, then model accuracy is improved, but computing resources and overhead increase
Solution Approach 1:
The patent applies partial action by using small image sets for frequent updates, which contains only the necessary subset of data needed to capture new features. This partial updating approach reduces computing overhead compared to using complete large image sets for every update, while still maintaining model accuracy through periodic comprehensive updates
Solution Approach 2:
The patent changes the parameter of image set size dynamically based on update frequency and model performance needs. Small image sets are used for frequent updates to minimize computing overhead, while large image sets are used periodically when substantial accuracy improvements are needed, optimizing the balance between resource usage and model performance
3Productivity
If the model is updated using small image sets, then retraining time is reduced, but the model's ability to capture new features is limited
Solution Approach 1:
The patent applies preliminary action by pre-selecting and curating small image sets that are specifically targeted at capturing new features. These pre-chosen images are designed to maximize feature capture capability within the constraints of fast updating, allowing the model to adapt quickly to new features without requiring extensive retraining
Data Source
AI summary
The present disclosure relates to a method, a device, and a computer program product for model updating. The method includes: acquiring a first image set and first annotation information, wherein the first annotation information indicates whether a corresponding image in the first image set includes a target object; updating a first version of an object verification model using the first image set and the first annotation information to obtain a second version, wherein the first version of the object verification model has been deployed to determine whether an input image includes the target object; determining the accuracy of the second version of the object verification model; and updating, if it is determined that the accuracy is lower than a preset accuracy threshold, the second version of the object verification model using a second image set and second annotation information to obtain a third version of the object verification model.