Inference Image Quality Alignment for Higher Model Accuracy

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The inference accuracy of inference processing for an input inference image is dependent on the quality of teacher images used for learning, and adjusting sensor operations based on inference results alone is insufficient to improve accuracy.

Innovation Solution

An information processing device and method that corrects the image quality of an input inference image based on the quality of teacher images used for learning, using parameters and pre-processing techniques to align the image quality with that of the teacher images.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If sensor operation is adjusted based on inference results, then inference processing can be optimized, but inference accuracy cannot be improved because teacher image quality is fixed

Engineering Contradiction:

Improveinference accuracyVSAvoidimage quality correction system

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The system performs preliminary action by correcting the image quality of the inference image before it is input to the inference model. The processing unit corrects the image quality based on pre-stored teacher image quality information, ensuring the inference image matches the expected input quality distribution of the model, thereby improving inference accuracy without requiring complex real-time adjustments during inference.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The processing unit acts as an intermediary between the sensor and the inference model. It receives the raw inference image from the sensor, corrects its image quality using teacher image quality information, and then supplies the corrected image to the inference model. This intermediary correction step bridges the gap between variable sensor output and fixed model expectations.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If teacher image quality is used to correct inference image, then inference accuracy improves, but additional processing steps and stored information are required

Engineering Contradiction:

Improveinference accuracyVSAvoidstored information

Core Design Contradiction:

ReliabilityVSQuantity of substance

Solution Approach 1:

The system uses copying by storing quality information (such as histograms or statistical characteristics) of teacher images in advance. Instead of storing the actual teacher images themselves, the system copies and stores their quality attributes, which are then used to correct inference images. This reduces the quantity of stored information while maintaining the ability to improve inference accuracy.

Inventive Principle:

Principle #26Copying

Data Source

PatentUS20250384681A1Information processing device, information processing method, and program

Publication Date: 2025.12.18 SONY SEMICON SOLUTIONS CORP

AI summary

The present technology relates to an information processing device, an information processing method, and a program that makes it possible to improve inference accuracy of inference processing for an inference image to be input. Inference processing is performed on an input inference image, and an image quality of the inference image is corrected based on an image quality of a teacher image used for learning in an inference unit.