Deep Learning Pose Transformation for Product Search Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Image-based product search techniques face accuracy issues due to variations in model pose, as the pose of a model (human or dummy) in an image affects the extraction of feature vectors, leading to unreliable product identification in online shopping malls.
Innovation Solution
A deep learning-based method that transforms the model pose in an input image into a standardized pose using a first standard pose, enabling consistent feature vector extraction and improving the accuracy of product searches by minimizing the impact of model pose variations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image-based product search is performed using existing search methods, then the search process is simple, but the search accuracy is degraded when the model pose varies
Solution Approach 1:
The system performs preliminary pose estimation and transformation before feature extraction. By pre-standardizing the model pose to a canonical position, the system prepares the image data in advance to ensure consistent feature vector extraction, thereby improving search accuracy without adding complexity during the actual search process
Solution Approach 2:
The patent introduces an intermediary pose transformation module that acts as a mediator between the input image and the feature extraction system. This intermediary component standardizes the model pose to a canonical position, serving as a bridge that reconciles the variability in input images with the requirements for consistent feature extraction, thus improving accuracy while maintaining system manageability
2Adaptability or versatility
If feature vector extraction is performed on images with varied model poses, then the system can handle diverse input images, but the feature vector reliability is degraded
Solution Approach 1:
The system changes the spatial parameters of the model pose by transforming images to a canonical pose representation. This parameter transformation standardizes the orientation and position of the model, ensuring that feature vectors are extracted under consistent geometric conditions, thereby improving reliability while maintaining the ability to process diverse input images through the transformation process
3Measurement precision
If pose standardization is performed using deep learning, then feature vector extraction accuracy is improved, but the processing time increases
Solution Approach 1:
The pose standardization and deep learning model training are performed in advance during system initialization or offline processing. By pre-computing the transformation parameters and feature extraction models, the system reduces the computational burden during actual image processing, thereby improving accuracy while minimizing real-time processing time
Solution Approach 2:
The patent replaces manual or traditional image processing methods with deep learning-based automated pose estimation and transformation. This substitution of mechanical processing with intelligent algorithms improves feature vector extraction accuracy while the automated nature of the process optimizes processing efficiency, reducing time loss compared to manual analysis
Data Source
AI summary
A method and system for performing a deep learning based product search obtain an input image including a target product to be searched; transform a model pose included in the input image; obtain a standard input image having the transformed pose of the model; obtain a main product image having an area including the target product by performing deep learning based on the standard input image; extract a feature vector from the main product image; perform a product search for a product similar to the target product based on the feature vector; and output a result of the product search.