Semantic Segmentation Graph Image Generation for Robust Training Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing models, particularly semantic segmentation models, face challenges in achieving high accuracy due to insufficient and varied real-world training datasets, leading to poor performance and low accuracy, as they often rely on limited image generation methods that do not account for complex environmental conditions.
Innovation Solution
A method involving semantic segmentation graphs and key words is employed to generate target images with specific features, ensuring semantic consistency and enhancing the training dataset, using a trained neural network model to map images and key words to a predetermined feature space, thereby improving model accuracy and robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If traditional image generation methods are used, then the training dataset can be expanded, but the accuracy and robustness of the image processing model remain insufficient due to limited variety and failure to account for complex environmental conditions
Solution Approach 1:
The patent segments the image generation process into distinct components: semantic segmentation graph generation, keyword-based feature transformation, and target image synthesis. This segmentation allows each component to be optimized independently, ensuring both dataset expansion and maintenance of high processing accuracy through controlled transformation of semantic features.
Solution Approach 2:
The patent transforms the semantic segmentation graph by applying parameter changes based on keywords representing environmental conditions (e.g., lighting, weather, season). These parameter changes generate diverse training images while maintaining semantic consistency, thereby improving model accuracy across varying environmental conditions without sacrificing dataset quantity.
2Reliability
If more diverse training data is generated to improve model robustness, then the model can handle complex environmental conditions better, but the complexity of the image generation process increases
Solution Approach 1:
The patent introduces a semantic segmentation graph as an intermediary representation between the source image and the final target image. This intermediary structure simplifies the generation process by working with segmented semantic features rather than raw pixels, reducing computational complexity while enabling diverse environmental condition transformations through keyword-based operations.
3Manufacturing precision
If semantic segmentation and transformation methods are used, then the authenticity and semantic consistency of generated images are maintained, but the computational complexity and processing time increase
Solution Approach 1:
The patent performs semantic segmentation and graph generation as preliminary actions before the actual image transformation. By pre-processing the source image into a semantic segmentation graph and preparing keyword-based transformation rules in advance, the system reduces the computational burden during runtime, maintaining high generation accuracy while reducing processing time for generating diverse training images.
Data Source
AI summary
Embodiments of the present disclosure relate to a method, an electronic device, and a computer program product for generating an image. The method includes acquiring a semantic segmentation graph by performing semantic segmentation on a source image. The method further includes acquiring a key word for describing a feature of a to-be-generated target image. The method further includes transforming the semantic segmentation graph by using the key word so as to acquire a transformed semantic segmentation graph. The method further includes generating the target image based on the transformed semantic segmentation graph. According to the method of embodiments of the present disclosure, a semantic segmentation graph of a source image and a key word can be used to generate a target image, so as to make the generated target image have a target feature and have semantic consistency with the source image, thereby generating a high-quality target image.