Feature Tensor Refinement Using Shared Statistical Parameters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video processing techniques are inefficient for machine vision-based systems and applications, as they are optimized for human video and image consumption.
Innovation Solution
A decoding device refines feature tensors by deriving statistical parameters from a bitstream, restoring and refining feature tensors using these parameters, and performing operations like analytics or inference with the refined tensors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional video processing techniques are used, then human video and image consumption is optimized, but machine vision-based systems efficiency deteriorates
Solution Approach 1:
The patent changes the parameters of video processing by introducing statistical parameter refinement (mean and standard deviation adjustment) to feature tensors. This transforms conventional video processing parameters to better suit machine vision requirements, enabling efficient machine-based analysis while maintaining compatibility with existing video coding frameworks.
Solution Approach 2:
The patent implements a feedback mechanism where statistical parameters are determined from decoded feature tensors and used to refine subsequent feature tensors. This iterative feedback process continuously improves the quality of feature representations for machine vision tasks while working within the existing video processing pipeline.
2Quantity of substance
If feature tensors are compressed from original feature tensors, then data transmission is reduced, but feature tensor quality deteriorates
Solution Approach 1:
The patent applies preliminary action by determining statistical parameters (mean and standard deviation) from the original feature tensors before compression. These pre-computed statistical parameters are then used to refine the compressed feature tensors, preserving quality information that would otherwise be lost during compression.
Solution Approach 2:
The patent uses parameter changes by adjusting the statistical parameters (mean and standard deviation) of the compressed feature tensors to match the original feature tensor distribution. This parameter adjustment restores the quality and characteristics of the compressed data, making it suitable for machine vision applications.
Data Source
AI summary
Disclosed herein are systems, methods, and instrumentalities associated with the refinement of feature tensors. A decoding device as described herein may obtain, from a bitstream, multiple feature tensors associated with a picture, wherein the multiple feature tensors may be compressed from respective original feature tensors associated with the picture. The decoding device may determine a single set of statistical parameters associated with the original feature tensors, restore the multiple feature tensors obtained from the bitstream to derive a set of restored feature tensors, and refine the set of restored feature tensors based on the single set of statistical parameters. The decoding device may then perform operations associated with the picture using the refined set of restored feature tensors. An encoding device as described herein may determine multiple feature tensors associated with a picture, determine a single set of statistical parameters associated with the multiple feature tensors, and encode the multiple feature tensors and the single set of statistical parameters into a bitstream.