Virtual Clothing Change Using Pose-Guided Image Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing virtual clothing changing methods are limited by large errors when dressing humans with different body shapes or poses in the same clothing, leading to mismatches between the clothing area and poses, and rough treatment of the connection with body parts of humans.
Innovation Solution
A method and apparatus for virtual clothing changing that involves obtaining source and target images, processing them to extract portrait and pose information, and then changing the clothing of the target human instance in the target image to match the target clothing from the source image, using this information to achieve accurate and flexible clothing changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If simple clothing replacement is used, then the operation is simple, but the accuracy of clothing fitting is poor
Solution Approach 1:
The patent segments the clothing changing process into multiple components: extracting clothing region from source image, extracting human body region from target image, obtaining pose information, and performing adaptive transformation. This segmentation allows each component to be processed separately with appropriate algorithms, improving overall accuracy while maintaining operational simplicity.
Solution Approach 2:
The patent changes the parameters of the clothing region (scaling, rotation, translation) based on the pose information and body shape of the target human instance. By dynamically adjusting these parameters rather than using fixed replacement, the system achieves accurate fitting across different poses and body types while keeping the operation simple.
2Adaptability or versatility
If the same clothing is used for different body shapes or poses, then the clothing selection is simple, but the matching accuracy is poor
Solution Approach 1:
The patent makes the clothing transformation dynamic by using pose information to determine the transformation parameters. The clothing region is transformed based on the target human instance's pose, allowing the same clothing to adapt to different body shapes and poses accurately. This dynamic approach enables versatility across different scenarios while maintaining high matching accuracy.
Solution Approach 2:
The system uses pose information as feedback to adjust the clothing transformation. By analyzing the pose of the target human instance and using this information to guide the clothing region transformation, the system achieves accurate matching for different body shapes and poses without requiring multiple clothing options.
3Device complexity
If clothing area is simply scaled or rotated, then the processing is simple, but the connection with body parts is rough
Solution Approach 1:
The patent performs preliminary extraction of the clothing region and preliminary obtaining of pose information before performing the final transformation. This preliminary processing prepares the data in advance, allowing the subsequent transformation to focus on achieving high-quality connection with body parts rather than dealing with raw, unprocessed data.
Data Source
AI summary
The present disclosure provides a virtual clothing changing method, apparatus, electronic device and readable medium. The method comprises: obtaining a source image and a target image, the source image comprising target clothing associated with a first human instance, the target image comprising a second human instance; obtaining first portrait information and first pose information of the source image and second pose information of the target image, respectively, by processing the source image and the target image; obtaining a clothing changed image by changing clothing of the second human instance in the target image to the target clothing in the source image according to the first portrait information, the first pose information and the second pose information.