Image-Based Property Recommendations Using Visual Feature Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional travel planning systems require users to manually search for destinations, excursions, and itineraries separately, making it difficult to identify pertinent destinations and properties that meet individual needs and preferences efficiently, leading to increased network occupancy and processing power consumption.
Innovation Solution
A computing system using a machine learning model to analyze a visual output, extract features, and identify similar visual outputs from a database, generating recommendations and improved graphical user interfaces that depict these features, reducing the need for multiple searches and improving search efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If users manually search for destinations, excursions, and itineraries separately using conventional systems, then they can find travel information, but the process requires multiple separate searches, increasing network occupancy and processing power consumption
Solution Approach 1:
The patent combines multiple separate search functions (destinations, excursions, itineraries, properties) into a single integrated search system that processes visual inputs and returns comprehensive travel recommendations in one operation, eliminating the need for multiple sequential searches and reducing network occupancy and processing power consumption
Solution Approach 2:
The patent introduces a machine learning model as an intermediary that analyzes visual outputs and extracts features to generate travel recommendations, acting as a mediator between user input and search results to reduce the complexity and resource requirements of the search process
2Measurement precision
If users manually search for properties with specific features, then they can find matching properties, but they struggle to articulate needs and preferences in search queries, requiring multiple search attempts
Solution Approach 1:
The patent replaces the mechanical system of manual keyword-based searching with an automated machine learning-based visual analysis system that extracts features from images and automatically generates accurate property matches, eliminating the need for users to manually articulate their needs and preferences
Solution Approach 2:
The system performs self-service by automatically analyzing visual inputs, extracting relevant features, and generating search queries without requiring user intervention to articulate preferences, thereby improving matching accuracy while reducing search process complexity
3Loss of information
If users review initial property images to determine if properties meet their needs, then they can identify suitable properties, but they cannot recognize whether properties meet preferences based on initial images alone, requiring additional searches
Solution Approach 1:
The patent performs preliminary action by extracting and analyzing features from property images before presenting search results to users, preparing comprehensive property information in advance so that users can immediately determine if properties meet their preferences without requiring additional searches
Solution Approach 2:
The system implements feedback by analyzing visual outputs and using extracted features to generate improved search results that directly address user preferences, creating a feedback loop that enhances information delivery and improves search efficiency
Data Source
AI summary
A computing system includes a processing circuit having a processor coupled to a memory device. The processing circuit performs operations including receiving, during a search session, a search query comprising configurable parameters; providing, during the search session, an array of visual outputs that meet criteria defined by the parameters; receiving, from a user device, during the search session, a selection of a visual output from the array; extracting, using a machine learning model and based on a pixel composition of the visual output, a feature; identifying, using the machine learning model, one or more second visual outputs related to travel properties that share the feature; replacing a default image associated with at least one travel property with an image representing the feature; and providing, to the user device, during the search session, a second array comprising the one or more second visual outputs including the replaced image.