Webpage Illustration Frame Acquisition via Deep Learning Scanning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing webpage illustration processing methods fail to directly cut target illustrations from original natural scene images effectively for webpage editing, resulting in inefficiencies and high costs due to poor accuracy and inappropriate image size considerations.
Innovation Solution
A method involving acquiring a first image set, labeling regions of interest, using a scanning window to create a sample training set, establishing an illustration frame acquisition model, and cutting images to obtain a target webpage illustration using depth learning algorithms for image detection and classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If traditional SVM machine learning is used for image automatic points of interest selection, then the processing can be automated, but the accuracy is poor due to not using deep learning algorithms
Solution Approach 1:
The patent changes the algorithmic parameters from traditional SVM to deep learning models (ResNet, VGG, etc.), fundamentally improving the accuracy of image point selection while maintaining automation. This parameter change enables the system to achieve both high automation and high precision.
2Adaptability or versatility
If existing automatic background image selection methods are used, then illustrations can be provided for promotional articles, but the method cannot handle cases where original images need to be cut due to inappropriate image size
Solution Approach 1:
The patent segments the image processing task into multiple stages: first selecting background images using deep learning classification, then separately handling image size adjustments through cutting and resizing operations. This segmentation allows the system to adapt to both promotional article requirements and image size issues independently.
Solution Approach 2:
The patent implements dynamic image size adjustment by detecting the actual size of extracted illustrations and automatically resizing or cutting them to match required webpage dimensions. This dynamic adaptation enables the system to handle various image size scenarios flexibly.
3Measurement precision
If manual labeling and selection of images is performed, then accuracy can be ensured, but the time cost and economic costs are high
Solution Approach 1:
The patent implements self-service automation where the deep learning system automatically performs image selection, extraction, and size adjustment without requiring manual intervention. The system serves itself by using pre-trained models to complete tasks that would otherwise require human annotators and editors, significantly reducing time costs while maintaining accuracy.
Data Source
AI summary
The processing method comprises: acquiring a first image set corresponding to a product category; labelling at least one region of interest for each image in the first image set, wherein each region of interest is used for representing an object; acquiring a scanning window; using the scanning window to scan the regions of interest, acquiring scanning results, and placing the scanning results into a sample training set; establishing an illustration frame acquisition model by taking the first image set as input and taking the sample training set as output; acquiring an image to be processed, and using the illustration frame acquisition model to acquire a target webpage illustration frame corresponding to the image to be processed; and cutting, according to the target webpage illustration frame, the image to be processed so as to obtain a target webpage illustration.