Task Output Compression for Accurate Distributed Image Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression technologies face limitations in increasing compression ratios while maintaining high accuracy for performing specific tasks, as they typically compress the entire image without considering task-specific outputs.
Innovation Solution
A method and apparatus for distributed image data processing that involves performing machine learning on an original image to generate multiple task outputs, combining these outputs to extract a final output, and compressing this final output for transmission, while allowing for additional tasks like object tracking and pose estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If the entire image is compressed and transmitted using traditional image compression technology, then the compression ratio can be increased, but the accuracy for performing specific tasks deteriorates
Solution Approach 1:
The patent segments the image processing task into multiple independent task-specific outputs (e.g., object detection, segmentation, keypoint detection) that are processed separately. Each task output is compressed independently based on its specific requirements, allowing optimized compression ratios for each task while maintaining overall task accuracy.
Solution Approach 2:
The patent applies different compression quality levels to different task outputs based on their specific requirements. Critical tasks requiring high accuracy maintain higher quality, while less critical tasks can use higher compression ratios. This local quality approach ensures task accuracy is maintained where needed while achieving overall compression improvement.
2Productivity
If the entire image is compressed and transmitted, then compression is applied uniformly, but the efficiency for task-specific processing deteriorates
Solution Approach 1:
The patent extracts only the necessary task-specific information from the original image rather than compressing and transmitting the entire image. By extracting specific task outputs (object detections, segments, keypoints) and compressing only these extracted elements, the system improves processing efficiency while reducing unnecessary data transmission volume.
Solution Approach 2:
The patent applies compression selectively to partial task outputs rather than uniformly to the entire image. Each task output receives appropriate compression treatment based on its specific requirements, avoiding excessive compression on critical data while achieving overall efficiency improvement through partial selective compression.
Data Source
AI summary
Disclosed herein are a method and apparatus for distributed image data processing. The method for distributed image data processing includes performing machine learning on an original image to produce a plurality of different task outputs, combining the plurality of task outputs to extract at least one final output, and compressing the final output and transmitting the final output to a server.