Selective Image Transmission for AI Sorting Network Load Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing method of sending large volumes of image data to a cloud server for sorting increases network load and cost, delays accuracy improvement, and requires remote cooperation between locations.
Innovation Solution
An information processing device that retains a model to process image data locally, determines sorting results, and selectively transmits only necessary data to an external device, using a network to reduce communication load and improve sorting accuracy through local learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all image data is transmitted to the cloud server for sorting, then sorting accuracy can be improved through AI processing, but network load and communication costs increase significantly
Solution Approach 1:
The patent extracts only the necessary information (sorting results and selective image data) from the complete image data set, transmitting only what is needed to the cloud server. This eliminates the need to transmit entire image files for every article, significantly reducing network traffic while maintaining sorting accuracy through selective data transmission.
Solution Approach 2:
The system segments the data transmission process into two parts: local preprocessing at the production site that handles routine sorting, and cloud-based AI processing that handles complex cases. This segmentation allows the majority of sorting to be done locally with minimal network communication, while only problematic or uncertain cases are transmitted to the cloud.
2Measurement precision
If large volumes of image data are transmitted for AI model construction, then model sorting accuracy can be improved, but the time required to collect and process data increases
Solution Approach 1:
The system performs preliminary sorting and filtering at the production site before transmitting data to the cloud. By pre-processing image data locally and identifying only those cases that require cloud-based AI analysis, the system reduces the time needed for data collection and preparation, while still gathering sufficient data for model construction.
Solution Approach 2:
The local system at the production site performs self-service preprocessing of image data, automatically identifying and selecting only the necessary data for cloud transmission. This self-service approach eliminates manual data collection and preparation time, allowing the system to efficiently gather training data without human intervention.
3Ease of operation
If image data is transmitted from production site to remote management location, then remote cooperation is enabled, but communication resource consumption and costs increase
Solution Approach 1:
The system extracts only essential data (sorting results and selected image samples) for transmission to remote management locations, rather than transmitting complete image data sets. This extraction approach enables remote cooperation and monitoring while minimizing communication resource consumption and associated costs.
Solution Approach 2:
Instead of transmitting original large-volume image data, the system transmits compressed representations or selective copies of image data that contain the essential information needed for remote management. This copying approach maintains the ability to perform remote analysis while significantly reducing communication overhead.
Data Source
AI summary
Disclosed herein are method, device and computer program to facilitate collaboration between different places using a network while reducing the consumption of communication resources to reduce cost. An edge device having a storage unit, a determination unit and an output control unit. The storage unit holds an AI model into which image data are sequentially input and which outputs a sorting result of each of the products. The determination unit determines whether the sorting result output from the AI model in response to the input image data satisfies a predetermined condition. When the predetermined condition is satisfied, the output control unit outputs the image data identified as being a defective product, together with a first text data indicating the determination result; and when the predetermined condition is not satisfied, the output control unit outputs a second text data indicating the determination result.