Sample Image Acquisition for Game Currency Recognition Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for obtaining sample images for training recognition networks in game currency recognition tasks lack diversity and quality, affecting the precision and generalization capability of the networks.
Innovation Solution
A method and apparatus for acquiring sample images of stacked bodies with diverse item information, including attributes and stacking modes, and filtering images based on quality conditions to ensure rich and accurate data for training neural networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If current methods are used to obtain sample images, then the acquisition process is simple, but the quality and diversity of sample images are insufficient
Solution Approach 1:
The image acquisition process is divided into multiple stages: initial sample collection, quality assessment, and iterative supplementation. Each stage focuses on specific aspects of image quality (diversity, clarity, distribution), allowing systematic improvement without overwhelming complexity.
Solution Approach 2:
The method implements a feedback mechanism where the quality and distribution of acquired images are continuously assessed against predefined standards. Images that do not meet quality criteria are identified and supplemented through targeted acquisition, creating a closed-loop system that progressively improves sample image quality.
2Adaptability or versatility
If more diverse sample images are acquired, then the generalization capability of recognition networks improves, but the time and resources required for acquisition increase
Solution Approach 1:
The method performs preliminary assessment of image diversity and distribution before full-scale acquisition. By identifying gaps in the sample set early, the system can target specific missing scenarios, avoiding unnecessary acquisition time while ensuring comprehensive coverage for better generalization.
Solution Approach 2:
Instead of acquiring all possible image variations uniformly, the method applies partial action by focusing resources on acquiring only the specific image types and scenarios that are currently missing or underrepresented, achieving diversity efficiency.
3Measurement precision
If strict image quality conditions are enforced, then the accuracy of recognition networks improves, but the number of usable sample images decreases
Solution Approach 1:
The method dynamically adjusts acquisition parameters (such as image resolution, lighting conditions, stacking configurations) to meet quality thresholds. By changing these parameters systematically, the system maintains high accuracy requirements while maximizing the number of acceptable images through optimized capture settings.
Solution Approach 2:
When original images do not meet quality criteria, the method creates alternative versions by adjusting capture parameters or generating synthetic variations, effectively multiplying the usable sample set from limited high-quality sources while maintaining accuracy standards.
Data Source
AI summary
Embodiments of the present disclosure provide methods and apparatuses for obtaining sample images, and electronic devices. The method includes: acquiring images of stacked bodies, where the stacked bodies in different images have different item information, and the item information comprises: an attribute and a stacking mode of a stacked body; and taking an acquired image as a sample image when the acquired image meets an image quality condition.