Inter-Channel Reference Feature Encoding for Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression technologies are not optimized for machine learning tasks, as they focus on high-resolution, high-quality image processing for human vision, leading to inefficiencies in processing vast amounts of image data required for artificial intelligence services.
Innovation Solution
A feature encoding/decoding method and apparatus based on inter-channel reference of a coding structure, sub-sampling, and up-sampling of feature data, which determines the coding structure of a current channel by referencing a previously reconstructed or encoded channel, and transmits a bitstream generated by the feature encoding method or apparatus.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing image compression technology is used, then high-resolution and high-quality image processing is achieved, but encoding/decoding efficiency is insufficient for machine learning tasks
Solution Approach 1:
The patent changes the fundamental parameters of compression by introducing channel-specific coding structures and inter-channel reference mechanisms. Instead of uniform compression, each channel is encoded independently with reference to previously encoded channels, optimizing the compression parameters for machine learning tasks while maintaining quality
Solution Approach 2:
The patent segments the feature map into multiple channels and applies different coding structures to each channel. By dividing the data into channel-specific units and processing them independently with reference to previous channels, the system achieves both high quality and improved encoding efficiency
2Productivity
If inter-channel reference is applied to determine coding structure, then encoding efficiency is improved, but device complexity increases
Solution Approach 1:
The patent creates a universal coding structure that can be applied across all channels through inter-channel reference. The same reference mechanism works for all channels, simplifying the overall system architecture while improving efficiency through shared coding patterns
Solution Approach 2:
The patent performs preliminary encoding of reference channels before encoding current channels. By pre-processing and storing coding structures of previous channels, the system reduces the complexity of determining coding structures during the encoding process, as current channels can directly reference pre-computed structures
Data Source
AI summary
Provided are a feature encoding/decoding method and device, a recording medium in which a bitstream generated by the feature encoding method is stored, and a method for transmitting the bitstream. A feature decoding method according to the present disclosure comprises the steps of: determining an encoding structure of a current channel in a feature map; acquiring, on the basis of the encoding structure, a current block by splitting the current channel; and reconstructing the current block, wherein the encoding structure of the current channel can be determined on the basis of whether an inter-channel reference referring to an encoding structure of a channel, having been reconstructed for the current channel, is applied.