Style Embedding Retrieval for Complementary Object Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
It is challenging to find and navigate content that showcases complementary and/or related fashion, décor, and other objects due to the ever-changing trends and vast amount of accessible information, making it difficult to identify outfits, décor, and accessories that complement each other.
Innovation Solution
A machine learning system, such as a trained neural network, is used to generate style embeddings for query objects, allowing for the identification of complementary and/or matching objects based on their style, using a corpus of content items processed for style embedding vectors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If manual browsing and navigation of fashion content is used, then users can explore available content, but it becomes difficult to find and navigate through content showing various outfits, looks, and trends due to the sheer amount of accessible content
Solution Approach 1:
The patent replaces manual browsing (mechanical navigation through content) with an automated machine learning system that uses neural networks to generate style embeddings and automatically identify complementary objects, thereby resolving the difficulty of navigating through vast amounts of fashion content
Solution Approach 2:
The system enables self-service by allowing users to input a query object and automatically receive identified complementary objects without manual intervention in the matching process, the machine learning system autonomously performs the complex task of style analysis and object matching
2Adaptability or versatility
If comprehensive fashion content is made available, then users have access to various trends and styles, but it becomes difficult to identify outfits and accessories that complement each other
Solution Approach 1:
The patent transforms the complex visual and stylistic properties of fashion objects into numerical style embedding vectors, changing the parameter representation from qualitative visual characteristics to quantitative vector data that can be systematically compared and measured for complementarity
Solution Approach 2:
The patent introduces style embeddings as an intermediary representation between the query object and the corpus of objects. These embeddings serve as a mediator that captures stylistic properties and enables systematic comparison to identify complementary objects without direct visual analysis
3Productivity
If style embedding vectors are generated and compared to identify complementary objects, then efficient retrieval of matching objects is enabled, but the system complexity increases due to the need for machine learning models and processing
Solution Approach 1:
The patent applies preliminary action by pre-processing fashion objects into style embedding vectors and organizing them in a structured corpus before query execution. This preprocessing step enables efficient retrieval during actual use, as the complex transformation work is done in advance
Solution Approach 2:
The patent uses style embedding vectors as simplified copies or representations of the actual fashion objects. Instead of directly comparing complex visual data, the system works with these vector copies that capture essential stylistic properties, reducing computational complexity while maintaining effectiveness
Data Source
AI summary
Described are systems and methods for determining complementary and/or matching objects based on an input query object. The described systems and methods can generate an embedding representative of the provided object, which can be transformed to generate a style embedding by a trained system, such as a machine learning system. The style embedding can then be used to identify one or more complementary objects from a corpus of classified objects. Aspects of the present disclosure also relate to creation of the training dataset, as well as training the machine learning system.