CNN Model Training Apparatus Dynamic FPN Block Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
There is a trade-off between accuracy and execution speed in CNN models, particularly when stacking Feature Pyramid Network (FPN) blocks, which affects real-time applications.
Innovation Solution
A model training apparatus and method that calculates the number of FPN blocks required for object detection in a dataset and estimates a parameter value for the number of FPN blocks based on this calculation, allowing for optimized CNN model training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the number of FPN blocks in the CNN model is increased, then detection accuracy is improved, but execution speed deteriorates
Solution Approach 1:
The patent changes the parameter of the number of FPN blocks from a fixed value to a dynamically determined value based on object characteristics. By calculating the required number of FPN blocks for each object type in the dataset and estimating optimal parameter values, the system adapts the model architecture to achieve necessary accuracy while minimizing execution time, thus resolving the trade-off between accuracy and speed.
2Reliability
If the number of FPN blocks is increased to improve accuracy, then detection performance is improved, but model complexity increases
Solution Approach 1:
The patent determines the optimal number of FPN blocks as a parameter based on dataset analysis and object detection requirements. By calculating the minimum required number of FPN blocks for each object type and using this information to estimate the optimal model parameter, the system achieves necessary detection performance while avoiding unnecessary model complexity that would result from using a fixed large number of FPN blocks.
3Productivity
If the number of FPN blocks is reduced to improve execution speed, then processing time is reduced, but detection accuracy deteriorates
Solution Approach 1:
The patent performs preliminary analysis of the dataset to calculate the number of FPN blocks required for each object type before training the model. By pre-calculating these requirements and using them to estimate the optimal parameter value, the system determines the minimum necessary model complexity in advance, ensuring that the model has sufficient capacity for accurate detection while avoiding excessive complexity that would slow down execution.
4Loss of time
If the number of FPN blocks is reduced to simplify the model, then training time is reduced, but detection capability deteriorates
Solution Approach 1:
The patent changes the approach from using a fixed number of FPN blocks to dynamically determining the optimal number based on dataset characteristics and detection requirements. By calculating the required FPN blocks for each object type and estimating the optimal parameter, the system identifies the minimum model complexity needed to achieve necessary detection capability, thereby reducing training time without sacrificing detection performance.
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: calculate, for objects in a dataset used in the past to train a Convolutional Neural Network (CNN) model, the number of Feature Pyramid Network (FPN) blocks in the CNN model required to detect the object, estimate a parameter value of the number of FPN blocks in the CNN model based on the calculated number of FPN blocks for the objects; and train the CNN model by using the estimated parameter value.