Object Posture Estimation via Style Conversion and Key Point Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for estimating object posture in augmented reality (AR) systems require significant computing resources and time, leading to latency and degraded user experience due to the need for high-precision real-time processing and image rendering.
Innovation Solution
A method that determines key point information in an image, modifies it using a key point feature map, and estimates object posture through a refinement process, converting the input image to a preset image style to improve accuracy and reduce computational load by integrating image-content and image-style features using a content backbone network and renderer.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-precision real-time processing and image rendering are performed to estimate object posture, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The system performs image style conversion in advance to transform the input image into a target style image that matches the training data distribution. This preliminary action ensures that subsequent key point detection and posture estimation can be performed more efficiently with higher accuracy, as the model doesn't need to adapt to style variations during processing
Solution Approach 2:
The posture estimation process is divided into separate modules: image style conversion, key point detection, and posture estimation. By segmenting the process, each module can be optimized independently - the style conversion handles appearance variations while the key point detector focuses on geometric features, improving overall efficiency and accuracy
2Measurement precision
If extensive image rendering and processing are performed to improve object posture estimation accuracy, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The system extracts and separates the style information from the input image through style conversion, isolating the geometric and semantic content that is essential for posture estimation. By taking out the style component, the system reduces computational complexity in subsequent processing steps while maintaining or improving accuracy
Solution Approach 2:
The system changes the style parameter of the input image to match the target style through style conversion. This parameter change aligns the input data distribution with the training data, allowing the posture estimation model to operate more efficiently with fewer computational resources required
Data Source
AI summary
An apparatus and method of estimating an object posture are provided. A method includes determining key point information in an image to be processed, determining modified key point information based on a key point feature map corresponding to the key point information, and estimating an object posture of an object in the image to be processed, based on the modified key point information.