CNN Activation Masking for Similar Image Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image search technologies fail to discriminate between regions useful and not useful for similar image search, leading to masking of relevant features and preventing suitable feature amount acquisition.
Innovation Solution
A feature amount acquisition device that derives activation levels in a multi-layer classifier to differentiate between high and low activation level regions in an image, using alpha blending for masking processing to emphasize relevant features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If masking processing is performed to remove background noise in image search, then search precision is improved, but useful regions other than the target object are also masked and feature amounts suitable for similar image search cannot be acquired
Solution Approach 1:
The patent applies local quality by differentiating masking intensity across different image regions based on their activation levels. Regions with high activation levels (important for classification) are preserved with minimal masking, while regions with low activation levels are masked more strongly. This selective, localized approach to masking ensures that useful feature information is retained while still removing irrelevant background noise, thereby resolving the contradiction between improving search precision and preserving useful feature information.
2Object-affected harmful factors
If faithful masking of non-target regions is performed, then background noise is removed, but discrimination between useful and non-useful regions is lost
Solution Approach 1:
The patent employs parameter changes by utilizing activation level values as a continuous parameter to control masking intensity. Instead of binary masking (fully masked or fully preserved), the system varies the masking parameter based on activation levels, allowing dynamic adjustment of masking strength. This enables discrimination between useful and non-useful regions while still removing background noise, as the masking parameter adapts to the importance of each region.
3Productivity
If all regions except target object are masked, then search efficiency is improved, but feature amounts for similar image search are reduced
Solution Approach 1:
The patent applies partial action by performing masking selectively rather than uniformly across all non-target regions. Instead of applying excessive masking to all areas outside the target object, the system applies partial masking only to regions with low activation levels, while preserving regions with high activation levels. This approach maintains search efficiency by removing irrelevant noise while preserving sufficient feature amounts for similar image search through selective preservation of important regions.
Data Source
AI summary
A feature amount acquisition device (100) includes an activation level calculator (12) deriving, as an activation level, a level at which, in a CNN classifier (11) including a plurality of layers and configured to, by processing input data based on image data of an input image capturing a first target and a second target around the first target in the layers, output a classification result of the first target, an unit in a layer among the layers influences a classification result and a feature amount acquirer (14) acquiring, based on the derived activation level and the image data of the input image, a feature amount of the input image so that a feature amount of a low activation level image region that is a region in the input image corresponding to a second unit having a lower activation level than a first unit is smaller than a feature amount of a high activation level image region that is a region in the input image corresponding to the first unit.