Automated Logo Indexing via Image Clustering and Query Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for visual recognition of logos within images are inefficient due to the labor-intensive nature of manual data collection and the lack of applicability of existing data mining techniques, which are not suited for logos unlike other object categories such as faces or landmarks.
Innovation Solution
A method involving generating a query list of logo search queries, clustering image search results, extracting representative images and names, and creating a logo index to identify logos within images, utilizing techniques such as image template generation, similarity scoring, and duplicate query merging to automate logo recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual or human-supervised collection of logo names and appearance models is used, then the accuracy of logo recognition is improved, but the labor intensity and time consumption increase significantly
Solution Approach 1:
The system performs self-learning by automatically extracting logo names and appearance models from image search results without human intervention. The automated pipeline includes generating query lists, retrieving image results, clustering images, extracting representative images, and building logo indices, all executed autonomously to eliminate manual data collection while maintaining recognition accuracy
Solution Approach 2:
The patent replaces manual mechanical data collection processes with automated computational systems. Manual supervision and human annotation are substituted by algorithmic processes including image template generation, similarity scoring, clustering algorithms, and automated logo index construction, thereby reducing time consumption while preserving measurement precision
2Extent of automation
If existing data mining techniques are applied to logos, then the automation of logo recognition is improved, but the applicability and effectiveness deteriorate due to differences between logos and other object categories
Solution Approach 1:
The system adapts data mining techniques to the specific local characteristics of logos by implementing logo-specific processing steps. This includes generating logo-specific query lists from search logs, extracting representative images based on logo appearance patterns, and building logo indices tailored to logo recognition requirements, thereby maintaining reliability while achieving automation
Solution Approach 2:
Instead of applying general object detection methods to logos, the patent inverts the approach by using logo-specific knowledge (from search queries and image results) to build specialized recognition models. The system leverages the unique properties of logos such as their appearance in search results and their distinct visual characteristics to create an inverted specialized system rather than a generic one
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for automatically extracting logos from images. Methods include generating a query list including a plurality of logo search queries, for each logo search query of the plurality of logo search queries: generating a plurality of image search results, each image search result including image data, and clustering the plurality of image search results into a plurality of clusters, each cluster including a plurality of images of the plurality of image search results, extracting, for each cluster of the plurality of clusters, a representative image to provide a plurality of representative images, and a name corresponding to the representative image to provide a plurality of names, and providing the plurality of representative images and the plurality of names to a logo index, the logo index being accessible to identify one or more logo images in a query image.