Image Analysis Device Using Probability Data for Convolutional Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Image analysis using reduced data with convolutional neural networks often results in degraded analysis accuracy due to the loss of image details.
Innovation Solution
An image analysis device that utilizes reduced image data as channel data for K channels and probability data as channel data for additional channels, improving analysis accuracy by incorporating probabilistic information alongside reduced image data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If reduced image data is used for image analysis, then processing speed is improved, but analysis accuracy is degraded
Solution Approach 1:
The system performs preliminary processing by generating probability data that indicates the likelihood of each pixel belonging to a specific object class. This probability data is generated in advance and stored, allowing the main image analysis to proceed quickly using reduced image data while incorporating the pre-computed probability information to maintain accuracy.
Solution Approach 2:
The system combines reduced image data with probability data to form a composite input for the convolutional neural network. This composite approach integrates the speed benefits of reduced image data with the accuracy benefits of probability-based object information, resolving the contradiction between processing speed and analysis accuracy.
2Quantity of substance
If reduced image data is used for image analysis, then memory consumption is reduced, but analysis accuracy is degraded
Solution Approach 1:
The system extracts essential object information from the full-resolution image by generating probability data that represents the likelihood of each pixel belonging to a specific object class. This extracted probability data is then used in conjunction with reduced image data, allowing the system to maintain analysis accuracy while working with smaller data volumes in memory.
Solution Approach 2:
By combining reduced image data with probability data, the system creates a composite representation that maintains the essential information needed for accurate analysis while significantly reducing the total memory requirements compared to using full-resolution images.
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
An image analysis device may obtain target image data representing a target image which is an analysis target, specify (m×n) partial images sequentially by scanning the target image data, wherein the (m×n) partial images are constituted of m partial images aligned along a first direction and n partial images aligned along a second direction, generate first probability data by using the (m×n) partial images and the first object data in the memory, and reduce the target image data so as to generate reduced image data. The image analysis device may execute image analysis according to a convolutional neural network by using the reduced image data as K pieces of channel data corresponding to K channels and using the first probability data as one piece of channel data corresponding to one channel, and output a result of the image analysis.