Server Apparatus for Image Tagging via Probability Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition systems using deep learning require large datasets with diverse images that accurately reflect the target object, but current methods either result in images that are overly focused on the photographer's intention or fail to include the target object due to positional information limitations, leading to inefficient and costly human intervention.
Innovation Solution
A method where a server processor calculates the probability of an image containing a specific target object and adds or requests information based on thresholds, using a combination of deep learning, histogram of oriented gradients (HOG), and support vector machines (SVM) to determine and tag images, allowing for efficient and accurate collection of diverse images without reflecting human intention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a large number of images are collected for deep learning, then recognition accuracy is improved, but the cost and time for manual tagging increases
Solution Approach 1:
The system applies partial automation by using probability thresholds to determine which images require manual tagging. Images with high probability matches are automatically tagged, while only ambiguous cases require human intervention, thus reducing overall manual tagging time while maintaining accuracy
Solution Approach 2:
The system performs self-tagging by automatically calculating probabilities and adding tags to images that meet the confidence threshold, eliminating the need for manual tagging of clear cases and reducing human cost
2Quantity of substance
If photographers are requested to capture specific objects, then relevant images are obtained, but the images reflect photographer's intention and lack diversity
Solution Approach 1:
Instead of having photographers actively search for specific objects, the system inverts the approach by having photographers capture general images and then using AI to identify and tag objects within them. This passive collection method reduces photographer bias and increases image diversity
3Extent of automation
If positional information is used to determine target inclusion, then automated processing is enabled, but accuracy decreases when target position is not clear
Solution Approach 1:
The system changes the parameter from binary positional information to probabilistic confidence scores. By calculating probability values and comparing them against thresholds, the system maintains automated processing while improving accuracy for ambiguous cases where the target position is not clear
4Measurement precision
If all images are manually reviewed for tagging, then tagging accuracy is high, but processing efficiency is low
Solution Approach 1:
The system segments the image processing workflow into automated probability calculation and manual verification stages. By dividing images into high-confidence (automatically tagged) and low-confidence (manually reviewed) groups, the system achieves both high accuracy and efficient processing
Data Source
AI summary
An image transmitted, through a network, from any of at least one terminal having a function of capturing an image or obtaining an image from another device is obtained. A probability that the obtained image includes a certain imaging target is calculated. If the probability is higher than a first threshold, information indicating the certain imaging target is added to the image. If the probability is lower than a second threshold, the information indicating the certain imaging target is not to the image. If the probability is equal to or higher than the second threshold and if the probability is equal to or lower than the first threshold, the image and request reception information for requesting addition of the information is transmitted to the image to any of the at least one terminal through the network.