Intelligent Model Upgrading for Cross-Domain Image Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Intelligent models trained in one domain experience performance degradation when applied to a different domain, leading to a decrease in accuracy in detecting images.
Innovation Solution
A method and apparatus for upgrading an intelligent model by acquiring input data from a new domain, setting annotation information based on output data, and training the model using both original and updated sample data to enhance its domain generalization performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the intelligent model is applied to a new domain different from the training domain, then the model can be reused across domains, but the performance and accuracy of the model will be degraded
Solution Approach 1:
The system performs preliminary actions by acquiring input data from the new domain before actual detection tasks. This pre-acquisition and pre-processing of domain-specific data enables the model to adapt to new domains while maintaining detection accuracy, resolving the contradiction between reusability and performance degradation.
Solution Approach 2:
The system implements feedback mechanisms by using output data from the model's processing of new domain data to generate annotation information. This feedback loop allows the model to learn from its own performance on new domains and adjust accordingly, maintaining high accuracy while achieving domain generalization.
2Measurement precision
If manual annotation is used for training samples, then the annotation information can be accurately obtained, but the labor resources and time required increase significantly
Solution Approach 1:
The system enables self-service by using the model's own output data to generate annotation information for training samples. Instead of requiring manual annotation, the model processes input data and the resulting output data is automatically converted into annotation information, significantly reducing labor resources while maintaining annotation accuracy.
Solution Approach 2:
The feedback mechanism allows the model to use its own output data to create training annotations. This self-feedback loop eliminates the need for manual annotation by transforming the model's processing results into the very annotations needed for training, reducing time consumption while preserving precision.
Data Source
AI summary
A method and apparatus for upgrading an intelligent model, an electronic device and A non-transitory computer readable storage medium are provided. The method includes: acquiring input data belonging to a first domain, wherein the first domain is different from a second domain of a first intelligent model to be upgraded, and the first intelligent model is obtained through training based on first sample data which belongs to the second domain; inputting the input data to the first intelligent model, and acquiring output data corresponding to the input data, wherein the output data is outputted by the first intelligent model from processing the input data, and the output data includes a confidence value and target box information; and training the first intelligent model according to the first sample data and the output data to obtain a second intelligent model.