Edge-Cloud Inference Filtering for High-Volume Data Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The transmission of large data volumes, such as high-definition camera images, from edge devices to cloud servers leads to increased data transfer and computation load on the server side, resulting in prolonged processing times.
Innovation Solution
A processing system that utilizes an edge device to divide processing data into multiple pieces, perform initial inference using lightweight models, and transmit only data meeting certain conditions to a server device for further processing with high-accuracy models, reducing data transfer and computation load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all processing data is transmitted to the cloud side, then processing can be performed with sufficient accuracy, but transmission path pressure increases and processing time becomes long
Solution Approach 1:
The patent divides the processing data into multiple pieces and processes them in parallel on the edge device using lightweight models. Only the results from pieces meeting predetermined conditions are transmitted to the cloud, segmenting the data processing workflow between edge and cloud to reduce transmission volume and time.
Solution Approach 2:
The edge device performs preliminary processing and inference using lightweight models before transmitting data to the cloud. This preliminary action filters and prepares data in advance, ensuring only necessary data is transmitted to the cloud, thereby reducing transmission time and path pressure.
2Adaptability or versatility
If all processing data is transmitted to the cloud side, then comprehensive processing can be performed, but transmission path pressure increases
Solution Approach 1:
The patent extracts and transmits only the necessary processed data pieces from the edge device to the cloud, rather than transmitting all raw data. The determination unit selects only data pieces that meet predetermined conditions, extracting essential information and reducing transmission path pressure while maintaining processing capability.
Solution Approach 2:
The processing data is segmented into multiple pieces, and only the relevant segments are transmitted to the cloud. This segmentation allows the system to maintain comprehensive processing capability while significantly reducing the volume of data transmitted, thereby lowering transmission path pressure.
3Productivity
If edge device processes data with large computation load, then processing can be completed faster locally, but computation resources become insufficient
Solution Approach 1:
The patent segments the computation task into multiple parallel processing pieces that can be handled by lightweight models on the edge device. This segmentation allows the edge device to process data faster using limited resources, while complex computations are offloaded to the cloud only when necessary.
Solution Approach 2:
The edge device performs partial processing using lightweight models for data pieces that can be handled locally, avoiding the need for full-computation resources. Only data pieces requiring higher accuracy trigger cloud processing, optimizing resource usage between edge and cloud.
4Quantity of substance
If lightweight models are used on edge device, then computation resources are saved, but inference accuracy decreases
Solution Approach 1:
The patent segments the inference task across different models with different accuracy levels. Lightweight models on the edge device handle routine inference for data pieces, while high-accuracy models on the cloud handle critical data pieces. This segmentation allows the system to optimize computation resources while maintaining overall inference accuracy.
Solution Approach 2:
Different quality levels of inference are applied locally at different stages. The edge device uses lightweight models for local inference on most data pieces, while the cloud provides high-accuracy inference only when needed. This local quality differentiation optimizes computation resources while maintaining necessary accuracy where required.
Data Source
AI summary
A processing system is performed by using an edge device and a server device, wherein the edge device includes first processing circuitry configured to input divided data obtained by dividing processing data into a plurality of pieces to a corresponding first model among a plurality of first models, and cause inference in each of the first models to be executed, and output, to the server device, only the divided data for which it is determined that an inference result in the corresponding first model matches a predetermined condition among a plurality of pieces of the divided data, and the server device includes second processing circuitry configured to execute inference processing on the divided data output from the edge device by using a second model having a higher amount of computation than that of the first model.