Feature Region Isolation for AI Model Training Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing model training methods face inefficiencies and inaccuracies due to interference from background content during AI imitation learning, leading to over-fitting and increased data processing burdens.
Innovation Solution
A method that isolates feature regions within video frames corresponding to card images, generates feature vectors based on these regions, and inputs them into a model for training, thereby reducing interference and improving training efficiency and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the entire content image is used for model training, then the model can learn comprehensive features, but background interference causes over-fitting and reduces training efficiency
Solution Approach 1:
The patent segments the content image into multiple regions of interest (ROIs) that contain actual card content, separating these from background interference areas. By processing only the segmented ROI regions rather than the entire image, the model achieves focused learning on relevant features while excluding background noise, thus improving training efficiency without sacrificing accuracy.
Solution Approach 2:
The patent extracts and isolates the useful feature regions containing card information from the full content image. Through region extraction, only the necessary card content is fed into the model for training, removing background interference completely. This extraction approach directly addresses the over-fitting issue caused by background content while maintaining comprehensive card feature learning.
2Loss of information
If the entire content image is processed for training, then all visual information is utilized, but data processing burden increases
Solution Approach 1:
The patent divides the content image into distinct regions of interest, identifying which areas contain actionable card information and which are background. By segmenting the data, the system processes only the necessary portions, reducing overall data processing volume while preserving all relevant card information needed for training.
Solution Approach 2:
The patent extracts only the essential card content regions from the full image, removing unnecessary background data. This extraction maintains complete card information for training purposes while significantly reducing the quantity of data that needs processing, thus lowering computational burden.
Data Source
AI summary
A processing circuitry obtains a training set based on an interaction process of a plurality of cards. The training set includes a plurality of video frames, and a video frame includes a trigger region for triggering an action during the interaction process. The processing circuitry determines feature regions in the trigger region. A feature region includes a card and is set with an action label for indicating a first training content in the feature region. The processing circuitry also determines a feature vector based on the plurality of video frames. The feature vector indicates a triggering of the feature region. The processing circuit inputs the feature vector into a first model for training to obtain a second model. The first model associates the feature vector with the action label, and the second model is used for indicating a correspondence between a content of the card and the action label.