ACNN Model Training Optimizing FPN Block Count
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In Artificial Convolutional Neural Networks (ACNNs), increasing the number of Feature Pyramid Network (FPN) blocks improves accuracy but significantly increases execution time, creating a trade-off between accuracy and efficiency, making it challenging to determine the optimal FPN count parameter value efficiently.
Innovation Solution
A model training apparatus and method that estimates the optimal number of FPN blocks using a training dataset, allowing for time-efficient training of the ACNN model by minimizing the number of FPN blocks required while maintaining accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If more FPN blocks are added to the ACNN model, then accuracy is improved, but execution time increases
Solution Approach 1:
The patent changes the parameter of FPN block count from a fixed manual value to a dynamically estimated value based on training dataset characteristics. By estimating the optimal FPN count parameter according to dataset properties, the system achieves high accuracy while minimizing execution time, resolving the trade-off between accuracy improvement and time consumption.
2Measurement precision
If more FPN blocks are added to the ACNN model, then accuracy is improved, but device complexity increases
Solution Approach 1:
The patent optimizes the FPN count parameter by estimating it from training dataset characteristics rather than using a fixed large value. This parameter optimization reduces the number of FPN blocks needed, thereby reducing model complexity and computational resources while maintaining high accuracy through adaptive parameter selection.
Data Source
AI summary
In one aspect, a model training apparatus includes at least one memory storing instructions; and at least one processor configured to execute the instructions to:—estimate a value of the number of FPN blocks included in an ACNN model by using a training dataset, and train the ACNN model by using the estimated value.